Default high profile for x264 when possible

miniaudio supports various backends. This will help reducing code
surface

.github/workflows/ci.yaml

@@ -18,7 +18,7 @@ jobs:
       - name: Checkout
         uses: actions/checkout@v2
       - name: Setup Go
-        uses: actions/setup-go@v1
+        uses: actions/setup-go@v2
         with:
           go-version: ${{ matrix.go }}
       - name: Install dependencies
@@ -30,15 +30,17 @@ jobs:
             libvpx-dev \
             libx264-dev
       - name: go vet
-        run: go vet ./...
+        run: go vet $(go list ./... | grep -v mmal)
       - name: go build
-        run: go build ./...
+        run: go build $(go list ./... | grep -v mmal)
       - name: go build without CGO
         run: go build . pkg/...
         env:
           CGO_ENABLED: 0
       - name: go test
-        run: go test ./... -v -race
+        run: go test -v -race -coverprofile=coverage.txt -covermode=atomic $(go list ./... | grep -v mmal)
+      - uses: codecov/codecov-action@v1
+        if: matrix.go == '1.15'
       - name: go test without CGO
         run: go test . pkg/... -v
         env:
@@ -53,7 +55,7 @@ jobs:
       - name: Checkout
         uses: actions/checkout@v2
       - name: Setup Go
-        uses: actions/setup-go@v1
+        uses: actions/setup-go@v2
         with:
           go-version: ${{ matrix.go }}
       - name: Install dependencies
@@ -64,15 +66,15 @@ jobs:
             libvpx \
             x264
       - name: go vet
-        run: go vet ./...
+        run: go vet $(go list ./... | grep -v mmal)
       - name: go build
-        run: go build ./...
+        run: go build $(go list ./... | grep -v mmal)
       - name: go build without CGO
         run: go build . pkg/...
         env:
           CGO_ENABLED: 0
       - name: go test
-        run: go test ./... -v -race
+        run: go test -v -race $(go list ./... | grep -v mmal)
       - name: go test without CGO
         run: go test . pkg/... -v
         env:

CODEOWNERS

@@ -1 +0,0 @@
-* @lherman-cs @at-wat

README.md

@@ -1,6 +1,17 @@
-# mediadevices
-
-Go implementation of the [MediaDevices](https://developer.mozilla.org/en-US/docs/Web/API/MediaDevices) API.
+<h1 align="center">
+  <br>
+  Pion MediaDevices
+  <br>
+</h1>
+<h4 align="center">Go implementation of the <a href="https://developer.mozilla.org/en-US/docs/Web/API/MediaDevices">MediaDevices</a> API</h4>
+<p align="center">
+  <a href="https://pion.ly/slack"><img src="https://img.shields.io/badge/join-us%20on%20slack-gray.svg?longCache=true&logo=slack&colorB=brightgreen" alt="Slack Widget"></a>
+  <a href="https://github.com/pion/mediadevices/actions"><img src="https://github.com/pion/mediadevices/workflows/CI/badge.svg?branch=master" alt="Build status"></a> 
+  <a href="https://pkg.go.dev/github.com/pion/mediadevices"><img src="https://godoc.org/github.com/pion/mediadevices?status.svg" alt="GoDoc"></a>
+  <a href="https://codecov.io/gh/pion/mediadevices"><img src="https://codecov.io/gh/pion/mediadevices/branch/master/graph/badge.svg" alt="Coverage Status"></a>
+  <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-yellow.svg" alt="License: MIT"></a>
+</p>
+<br>
 
 ![](img/demo.gif)
 

codec.go

@@ -0,0 +1,117 @@
+package mediadevices
+
+import (
+	"errors"
+	"fmt"
+	"strings"
+
+	"github.com/pion/mediadevices/pkg/codec"
+	"github.com/pion/mediadevices/pkg/io/audio"
+	"github.com/pion/mediadevices/pkg/io/video"
+	"github.com/pion/mediadevices/pkg/prop"
+	"github.com/pion/webrtc/v2"
+)
+
+// CodecSelector is a container of video and audio encoder builders, which later will be used
+// for codec matching.
+type CodecSelector struct {
+	videoEncoders []codec.VideoEncoderBuilder
+	audioEncoders []codec.AudioEncoderBuilder
+}
+
+// CodecSelectorOption is a type for specifying CodecSelector options
+type CodecSelectorOption func(*CodecSelector)
+
+// WithVideoEncoders replace current video codecs with listed encoders
+func WithVideoEncoders(encoders ...codec.VideoEncoderBuilder) CodecSelectorOption {
+	return func(t *CodecSelector) {
+		t.videoEncoders = encoders
+	}
+}
+
+// WithVideoEncoders replace current audio codecs with listed encoders
+func WithAudioEncoders(encoders ...codec.AudioEncoderBuilder) CodecSelectorOption {
+	return func(t *CodecSelector) {
+		t.audioEncoders = encoders
+	}
+}
+
+// NewCodecSelector constructs CodecSelector with given variadic options
+func NewCodecSelector(opts ...CodecSelectorOption) *CodecSelector {
+	var track CodecSelector
+
+	for _, opt := range opts {
+		opt(&track)
+	}
+
+	return &track
+}
+
+// Populate lets the webrtc engine be aware of supported codecs that are contained in CodecSelector
+func (selector *CodecSelector) Populate(setting *webrtc.MediaEngine) {
+	for _, encoder := range selector.videoEncoders {
+		setting.RegisterCodec(encoder.RTPCodec().RTPCodec)
+	}
+
+	for _, encoder := range selector.audioEncoders {
+		setting.RegisterCodec(encoder.RTPCodec().RTPCodec)
+	}
+}
+
+func (selector *CodecSelector) selectVideoCodec(wantCodecs []*webrtc.RTPCodec, reader video.Reader, inputProp prop.Media) (codec.ReadCloser, *codec.RTPCodec, error) {
+	var selectedEncoder codec.VideoEncoderBuilder
+	var encodedReader codec.ReadCloser
+	var errReasons []string
+	var err error
+
+outer:
+	for _, wantCodec := range wantCodecs {
+		name := wantCodec.Name
+		for _, encoder := range selector.videoEncoders {
+			if encoder.RTPCodec().Name == name {
+				encodedReader, err = encoder.BuildVideoEncoder(reader, inputProp)
+				if err == nil {
+					selectedEncoder = encoder
+					break outer
+				}
+			}
+
+			errReasons = append(errReasons, fmt.Sprintf("%s: %s", encoder.RTPCodec().Name, err))
+		}
+	}
+
+	if selectedEncoder == nil {
+		return nil, nil, errors.New(strings.Join(errReasons, "\n\n"))
+	}
+
+	return encodedReader, selectedEncoder.RTPCodec(), nil
+}
+
+func (selector *CodecSelector) selectAudioCodec(wantCodecs []*webrtc.RTPCodec, reader audio.Reader, inputProp prop.Media) (codec.ReadCloser, *codec.RTPCodec, error) {
+	var selectedEncoder codec.AudioEncoderBuilder
+	var encodedReader codec.ReadCloser
+	var errReasons []string
+	var err error
+
+outer:
+	for _, wantCodec := range wantCodecs {
+		name := wantCodec.Name
+		for _, encoder := range selector.audioEncoders {
+			if encoder.RTPCodec().Name == name {
+				encodedReader, err = encoder.BuildAudioEncoder(reader, inputProp)
+				if err == nil {
+					selectedEncoder = encoder
+					break outer
+				}
+			}
+
+			errReasons = append(errReasons, fmt.Sprintf("%s: %s", encoder.RTPCodec().Name, err))
+		}
+	}
+
+	if selectedEncoder == nil {
+		return nil, nil, errors.New(strings.Join(errReasons, "\n\n"))
+	}
+
+	return encodedReader, selectedEncoder.RTPCodec(), nil
+}

codecov.yml

@@ -0,0 +1,10 @@
+coverage:
+  status:
+    project:
+      default:
+        # Allow decreasing 2% of total coverage to avoid noise.
+        threshold: 2%
+    patch: off
+
+ignore:
+  - "examples/*"

examples/facedetection/README.md

@@ -1,29 +0,0 @@
-## Instructions
-
-### Download facedetection
-
-```
-go get github.com/pion/mediadevices/examples/facedetection
-```
-
-### Open example page
-
-[jsfiddle.net](https://jsfiddle.net/gh/get/library/pure/pion/mediadevices/tree/master/examples/internal/jsfiddle/video) you should see two text-areas and a 'Start Session' button
-
-### Run facedetection with your browsers SessionDescription as stdin
-
-In the jsfiddle the top textarea is your browser, copy that and:
-
-#### Linux
-
-Run `echo $BROWSER_SDP | facedetection`
-
-### Input facedetection's SessionDescription into your browser
-
-Copy the text that `facedetection` just emitted and copy into second text area
-
-### Hit 'Start Session' in jsfiddle, enjoy your video!
-
-A video should start playing in your browser above the input boxes, and will continue playing until you close the application.
-
-Congrats, you have used pion-WebRTC! Now start building something cool

examples/facedetection/detector.go

@@ -1,118 +0,0 @@
-package main
-
-import (
-	"image"
-	"image/color"
-	"image/draw"
-	"io/ioutil"
-	"log"
-
-	"github.com/disintegration/imaging"
-	pigo "github.com/esimov/pigo/core"
-)
-
-var (
-	cascade    []byte
-	err        error
-	classifier *pigo.Pigo
-)
-
-func imgToGrayscale(img image.Image) []uint8 {
-	bounds := img.Bounds()
-	flatten := bounds.Dy() * bounds.Dx()
-	grayImg := make([]uint8, flatten)
-
-	i := 0
-	for y := bounds.Min.Y; y < bounds.Max.Y; y++ {
-		for x := bounds.Min.X; x < bounds.Max.X; x++ {
-			pix := img.At(x, y)
-			grayPix := color.GrayModel.Convert(pix).(color.Gray)
-			grayImg[i] = grayPix.Y
-			i++
-		}
-	}
-	return grayImg
-}
-
-// clusterDetection runs Pigo face detector core methods
-// and returns a cluster with the detected faces coordinates.
-func clusterDetection(img image.Image) []pigo.Detection {
-	grayscale := imgToGrayscale(img)
-	bounds := img.Bounds()
-	cParams := pigo.CascadeParams{
-		MinSize:     100,
-		MaxSize:     600,
-		ShiftFactor: 0.15,
-		ScaleFactor: 1.1,
-		ImageParams: pigo.ImageParams{
-			Pixels: grayscale,
-			Rows:   bounds.Dy(),
-			Cols:   bounds.Dx(),
-			Dim:    bounds.Dx(),
-		},
-	}
-
-	if len(cascade) == 0 {
-		cascade, err = ioutil.ReadFile("facefinder")
-		if err != nil {
-			log.Fatalf("Error reading the cascade file: %s", err)
-		}
-		p := pigo.NewPigo()
-
-		// Unpack the binary file. This will return the number of cascade trees,
-		// the tree depth, the threshold and the prediction from tree's leaf nodes.
-		classifier, err = p.Unpack(cascade)
-		if err != nil {
-			log.Fatalf("Error unpacking the cascade file: %s", err)
-		}
-	}
-
-	// Run the classifier over the obtained leaf nodes and return the detection results.
-	// The result contains quadruplets representing the row, column, scale and detection score.
-	dets := classifier.RunCascade(cParams, 0.0)
-
-	// Calculate the intersection over union (IoU) of two clusters.
-	dets = classifier.ClusterDetections(dets, 0)
-
-	return dets
-}
-
-func drawCircle(img draw.Image, x0, y0, r int, c color.Color) {
-	x, y, dx, dy := r-1, 0, 1, 1
-	err := dx - (r * 2)
-
-	for x > y {
-		img.Set(x0+x, y0+y, c)
-		img.Set(x0+y, y0+x, c)
-		img.Set(x0-y, y0+x, c)
-		img.Set(x0-x, y0+y, c)
-		img.Set(x0-x, y0-y, c)
-		img.Set(x0-y, y0-x, c)
-		img.Set(x0+y, y0-x, c)
-		img.Set(x0+x, y0-y, c)
-
-		if err <= 0 {
-			y++
-			err += dy
-			dy += 2
-		}
-		if err > 0 {
-			x--
-			dx += 2
-			err += dx - (r * 2)
-		}
-	}
-}
-
-func markFaces(img image.Image) image.Image {
-	nrgba := imaging.Clone(img)
-	dets := clusterDetection(img)
-	for _, det := range dets {
-		if det.Q < 5.0 {
-			continue
-		}
-
-		drawCircle(nrgba, det.Col, det.Row, det.Scale/2, color.Black)
-	}
-	return nrgba
-}

examples/facedetection/main.go

@@ -1,119 +0,0 @@
-package main
-
-import (
-	"fmt"
-	"image"
-
-	"github.com/pion/mediadevices"
-	"github.com/pion/mediadevices/examples/internal/signal"
-	"github.com/pion/mediadevices/pkg/codec"
-	"github.com/pion/mediadevices/pkg/codec/vpx"       // This is required to use VP8/VP9 video encoder
-	_ "github.com/pion/mediadevices/pkg/driver/camera" // This is required to register camera adapter
-	"github.com/pion/mediadevices/pkg/frame"
-	"github.com/pion/mediadevices/pkg/io/video"
-	"github.com/pion/mediadevices/pkg/prop"
-	"github.com/pion/webrtc/v2"
-)
-
-func markFacesTransformer(r video.Reader) video.Reader {
-	return video.ReaderFunc(func() (img image.Image, err error) {
-		img, err = r.Read()
-		if err != nil {
-			return
-		}
-
-		img = markFaces(img)
-		return
-	})
-}
-
-func main() {
-	config := webrtc.Configuration{
-		ICEServers: []webrtc.ICEServer{
-			{
-				URLs: []string{"stun:stun.l.google.com:19302"},
-			},
-		},
-	}
-
-	// Wait for the offer to be pasted
-	offer := webrtc.SessionDescription{}
-	signal.Decode(signal.MustReadStdin(), &offer)
-
-	// Create a new RTCPeerConnection
-	mediaEngine := webrtc.MediaEngine{}
-	if err := mediaEngine.PopulateFromSDP(offer); err != nil {
-		panic(err)
-	}
-	api := webrtc.NewAPI(webrtc.WithMediaEngine(mediaEngine))
-	peerConnection, err := api.NewPeerConnection(config)
-	if err != nil {
-		panic(err)
-	}
-
-	// Set the handler for ICE connection state
-	// This will notify you when the peer has connected/disconnected
-	peerConnection.OnICEConnectionStateChange(func(connectionState webrtc.ICEConnectionState) {
-		fmt.Printf("Connection State has changed %s \n", connectionState.String())
-	})
-
-	md := mediadevices.NewMediaDevices(peerConnection)
-
-	vp8Params, err := vpx.NewVP8Params()
-	if err != nil {
-		panic(err)
-	}
-	vp8Params.BitRate = 100000 // 100kbps
-
-	s, err := md.GetUserMedia(mediadevices.MediaStreamConstraints{
-		Video: func(c *mediadevices.MediaTrackConstraints) {
-			c.FrameFormat = prop.FrameFormatExact(frame.FormatI420) // most of the encoder accepts I420
-			c.Enabled = true
-			c.Width = prop.Int(640)
-			c.Height = prop.Int(480)
-			c.VideoTransform = markFacesTransformer
-			c.VideoEncoderBuilders = []codec.VideoEncoderBuilder{&vp8Params}
-		},
-	})
-	if err != nil {
-		panic(err)
-	}
-
-	for _, tracker := range s.GetTracks() {
-		t := tracker.Track()
-		tracker.OnEnded(func(err error) {
-			fmt.Printf("Track (ID: %s, Label: %s) ended with error: %v\n",
-				t.ID(), t.Label(), err)
-		})
-		_, err = peerConnection.AddTransceiverFromTrack(t,
-			webrtc.RtpTransceiverInit{
-				Direction: webrtc.RTPTransceiverDirectionSendonly,
-			},
-		)
-		if err != nil {
-			panic(err)
-		}
-	}
-
-	// Set the remote SessionDescription
-	err = peerConnection.SetRemoteDescription(offer)
-	if err != nil {
-		panic(err)
-	}
-
-	// Create an answer
-	answer, err := peerConnection.CreateAnswer(nil)
-	if err != nil {
-		panic(err)
-	}
-
-	// Sets the LocalDescription, and starts our UDP listeners
-	err = peerConnection.SetLocalDescription(answer)
-	if err != nil {
-		panic(err)
-	}
-
-	// Output the answer in base64 so we can paste it in browser
-	fmt.Println(signal.Encode(answer))
-	select {}
-}

examples/go.mod

@@ -2,8 +2,8 @@ module github.com/pion/mediadevices/examples
 
 go 1.14
 
-replace github.com/pion/mediadevices => ../
-
 // Please don't commit require entries of examples.
 // `git checkout master examples/go.mod` to revert this file.
-require github.com/pion/mediadevices v0.0.0-00010101000000-000000000000
+require github.com/pion/mediadevices v0.0.0
+
+replace github.com/pion/mediadevices v0.0.0 => ../

examples/http/main.go

@@ -0,0 +1,77 @@
+// This is an example of using mediadevices to broadcast your camera through http.
+// The example doesn't aim to be performant, but rather it strives to be simple.
+package main
+
+import (
+	"bytes"
+	"fmt"
+	"image/jpeg"
+	"io"
+	"log"
+	"mime/multipart"
+	"net/http"
+	"net/textproto"
+
+	"github.com/pion/mediadevices"
+	"github.com/pion/mediadevices/pkg/prop"
+
+	// Note: If you don't have a camera or microphone or your adapters are not supported,
+	//       you can always swap your adapters with our dummy adapters below.
+	// _ "github.com/pion/mediadevices/pkg/driver/videotest"
+	_ "github.com/pion/mediadevices/pkg/driver/camera" // This is required to register camera adapter
+)
+
+func must(err error) {
+	if err != nil {
+		panic(err)
+	}
+}
+
+func main() {
+	s, err := mediadevices.GetUserMedia(mediadevices.MediaStreamConstraints{
+		Video: func(constraint *mediadevices.MediaTrackConstraints) {
+			constraint.Width = prop.Int(600)
+			constraint.Height = prop.Int(400)
+		},
+	})
+	must(err)
+
+	t := s.GetVideoTracks()[0]
+	videoTrack := t.(*mediadevices.VideoTrack)
+
+	http.HandleFunc("/", func(w http.ResponseWriter, r *http.Request) {
+		var buf bytes.Buffer
+		videoReader := videoTrack.NewReader(false)
+		mimeWriter := multipart.NewWriter(w)
+
+		contentType := fmt.Sprintf("multipart/x-mixed-replace;boundary=%s", mimeWriter.Boundary())
+		w.Header().Add("Content-Type", contentType)
+
+		partHeader := make(textproto.MIMEHeader)
+		partHeader.Add("Content-Type", "image/jpeg")
+
+		for {
+			frame, release, err := videoReader.Read()
+			if err == io.EOF {
+				return
+			}
+			must(err)
+
+			err = jpeg.Encode(&buf, frame, nil)
+			// Since we're done with img, we need to release img so that that the original owner can reuse
+			// this memory.
+			release()
+			must(err)
+
+			partWriter, err := mimeWriter.CreatePart(partHeader)
+			must(err)
+
+			_, err = partWriter.Write(buf.Bytes())
+			buf.Reset()
+			must(err)
+		}
+	})
+
+	fmt.Println("listening on http://localhost:1313")
+	log.Println(http.ListenAndServe("localhost:1313", nil))
+}

examples/rtp-send/README.md

@@ -1,29 +0,0 @@
-## Instructions
-
-### Download rtp-send example
-
-```
-go get github.com/pion/mediadevices/examples/rtp-send
-```
-
-### Listen RTP
-
-Install GStreamer and run:
-```
-gst-launch-1.0 udpsrc port=5000 caps=application/x-rtp,encode-name=VP8 \
-    ! rtpvp8depay ! vp8dec ! videoconvert ! autovideosink
-```
-
-Or run VLC media plyer:
-```
-vlc ./vp8.sdp
-```
-
-### Run rtp-send
-
-Run `rtp-send localhost:5000`
-
-A video should start playing in your GStreamer or VLC window.
-It's not WebRTC, but pure RTP.
-
-Congrats, you have used pion-MediaDevices! Now start building something cool

examples/rtp-send/main.go

@@ -1,120 +0,0 @@
-package main
-
-import (
-	"fmt"
-	"net"
-	"os"
-
-	"github.com/pion/mediadevices"
-	"github.com/pion/mediadevices/pkg/codec"
-	"github.com/pion/mediadevices/pkg/codec/vpx"       // This is required to use VP8/VP9 video encoder
-	_ "github.com/pion/mediadevices/pkg/driver/camera" // This is required to register camera adapter
-	"github.com/pion/mediadevices/pkg/frame"
-	"github.com/pion/mediadevices/pkg/prop"
-	"github.com/pion/rtp"
-	"github.com/pion/webrtc/v2"
-	"github.com/pion/webrtc/v2/pkg/media"
-)
-
-const (
-	mtu = 1000
-)
-
-func main() {
-	if len(os.Args) != 2 {
-		fmt.Printf("usage: %s host:port\n", os.Args[0])
-		return
-	}
-
-	md := mediadevices.NewMediaDevicesFromCodecs(
-		map[webrtc.RTPCodecType][]*webrtc.RTPCodec{
-			webrtc.RTPCodecTypeVideo: []*webrtc.RTPCodec{
-				webrtc.NewRTPVP8Codec(100, 90000),
-			},
-		},
-		mediadevices.WithTrackGenerator(
-			func(_ uint8, _ uint32, id, _ string, codec *webrtc.RTPCodec) (
-				mediadevices.LocalTrack, error,
-			) {
-				return newTrack(codec, id, os.Args[1]), nil
-			},
-		),
-	)
-
-	vp8Params, err := vpx.NewVP8Params()
-	if err != nil {
-		panic(err)
-	}
-	vp8Params.BitRate = 100000 // 100kbps
-
-	_, err = md.GetUserMedia(mediadevices.MediaStreamConstraints{
-		Video: func(c *mediadevices.MediaTrackConstraints) {
-			c.FrameFormat = prop.FrameFormat(frame.FormatYUY2)
-			c.Enabled = true
-			c.Width = prop.Int(640)
-			c.Height = prop.Int(480)
-			c.VideoEncoderBuilders = []codec.VideoEncoderBuilder{&vp8Params}
-		},
-	})
-	if err != nil {
-		panic(err)
-	}
-
-	select {}
-}
-
-type track struct {
-	codec      *webrtc.RTPCodec
-	packetizer rtp.Packetizer
-	id         string
-	conn       net.Conn
-}
-
-func newTrack(codec *webrtc.RTPCodec, id, dest string) *track {
-	addr, err := net.ResolveUDPAddr("udp", dest)
-	if err != nil {
-		panic(err)
-	}
-	conn, err := net.DialUDP("udp", nil, addr)
-	if err != nil {
-		panic(err)
-	}
-	return &track{
-		codec: codec,
-		packetizer: rtp.NewPacketizer(
-			mtu,
-			codec.PayloadType,
-			1,
-			codec.Payloader,
-			rtp.NewRandomSequencer(),
-			codec.ClockRate,
-		),
-		id:   id,
-		conn: conn,
-	}
-}
-
-func (t *track) WriteSample(s media.Sample) error {
-	buf := make([]byte, mtu)
-	pkts := t.packetizer.Packetize(s.Data, s.Samples)
-	for _, p := range pkts {
-		n, err := p.MarshalTo(buf)
-		if err != nil {
-			panic(err)
-		}
-		_, _ = t.conn.Write(buf[:n])
-	}
-	return nil
-}
-
-func (t *track) Codec() *webrtc.RTPCodec {
-	return t.codec
-}
-
-func (t *track) ID() string {
-	return t.id
-}
-
-func (t *track) Kind() webrtc.RTPCodecType {
-	return t.codec.Type
-}

examples/rtp-send/vp8.sdp

@@ -1,9 +0,0 @@
-v=0
-o=- 1234567890 1234567890 IN IP4 0.0.0.0
-s=RTP-Send Example
-i=Example
-c=IN IP4 0.0.0.0
-t=0 0
-a=recvonly
-m=video 5000 RTP/AVP 100
-a=rtpmap:100 VP8/90000

examples/screenshare/README.md

@@ -1,29 +0,0 @@
-## Instructions
-
-### Download screenshare
-
-```
-go get github.com/pion/mediadevices/examples/screenshare
-```
-
-### Open example page
-
-[jsfiddle.net](https://jsfiddle.net/gh/get/library/pure/pion/mediadevices/tree/master/examples/internal/jsfiddle/audio-and-video) you should see two text-areas and a 'Start Session' button
-
-### Run screenshare with your browsers SessionDescription as stdin
-
-In the jsfiddle the top textarea is your browser, copy that and:
-
-#### Linux
-
-Run `echo $BROWSER_SDP | screenshare`
-
-### Input screenshare's SessionDescription into your browser
-
-Copy the text that `screenshare` just emitted and copy into second text area
-
-### Hit 'Start Session' in jsfiddle, enjoy your video!
-
-A video should start playing in your browser above the input boxes, and will continue playing until you close the application.
-
-Congrats, you have used pion-WebRTC! Now start building something cool

examples/screenshare/main.go

@@ -1,101 +0,0 @@
-package main
-
-import (
-	"fmt"
-
-	"github.com/pion/mediadevices"
-	"github.com/pion/mediadevices/examples/internal/signal"
-	"github.com/pion/mediadevices/pkg/codec"
-	"github.com/pion/mediadevices/pkg/codec/vpx"       // This is required to use VP8/VP9 video encoder
-	_ "github.com/pion/mediadevices/pkg/driver/screen" // This is required to register screen capture adapter
-	"github.com/pion/mediadevices/pkg/io/video"
-	"github.com/pion/webrtc/v2"
-)
-
-func main() {
-	config := webrtc.Configuration{
-		ICEServers: []webrtc.ICEServer{
-			{
-				URLs: []string{"stun:stun.l.google.com:19302"},
-			},
-		},
-	}
-
-	// Wait for the offer to be pasted
-	offer := webrtc.SessionDescription{}
-	signal.Decode(signal.MustReadStdin(), &offer)
-
-	// Create a new RTCPeerConnection
-	mediaEngine := webrtc.MediaEngine{}
-	if err := mediaEngine.PopulateFromSDP(offer); err != nil {
-		panic(err)
-	}
-	api := webrtc.NewAPI(webrtc.WithMediaEngine(mediaEngine))
-	peerConnection, err := api.NewPeerConnection(config)
-	if err != nil {
-		panic(err)
-	}
-
-	// Set the handler for ICE connection state
-	// This will notify you when the peer has connected/disconnected
-	peerConnection.OnICEConnectionStateChange(func(connectionState webrtc.ICEConnectionState) {
-		fmt.Printf("Connection State has changed %s \n", connectionState.String())
-	})
-
-	md := mediadevices.NewMediaDevices(peerConnection)
-
-	vp8Params, err := vpx.NewVP8Params()
-	if err != nil {
-		panic(err)
-	}
-	vp8Params.BitRate = 100000 // 100kbps
-
-	s, err := md.GetDisplayMedia(mediadevices.MediaStreamConstraints{
-		Video: func(c *mediadevices.MediaTrackConstraints) {
-			c.Enabled = true
-			c.VideoTransform = video.Scale(-1, 360, nil) // Resize to 360p
-			c.VideoEncoderBuilders = []codec.VideoEncoderBuilder{&vp8Params}
-		},
-	})
-	if err != nil {
-		panic(err)
-	}
-
-	for _, tracker := range s.GetTracks() {
-		t := tracker.Track()
-		tracker.OnEnded(func(err error) {
-			fmt.Printf("Track (ID: %s, Label: %s) ended with error: %v\n",
-				t.ID(), t.Label(), err)
-		})
-		_, err = peerConnection.AddTransceiverFromTrack(t,
-			webrtc.RtpTransceiverInit{
-				Direction: webrtc.RTPTransceiverDirectionSendonly,
-			},
-		)
-		if err != nil {
-			panic(err)
-		}
-	}
-
-	// Set the remote SessionDescription
-	err = peerConnection.SetRemoteDescription(offer)
-	if err != nil {
-		panic(err)
-	}
-
-	// Create an answer
-	answer, err := peerConnection.CreateAnswer(nil)
-	if err != nil {
-		panic(err)
-	}
-
-	// Sets the LocalDescription, and starts our UDP listeners
-	err = peerConnection.SetLocalDescription(answer)
-	if err != nil {
-		panic(err)
-	}
-
-	// Output the answer in base64 so we can paste it in browser
-	fmt.Println(signal.Encode(answer))
-	select {}
-}

examples/webrtc/README.md

@@ -3,24 +3,24 @@
 ### Download gstreamer-send
 
 ```
-go get github.com/pion/mediadevices/examples/simple
+go get github.com/pion/mediadevices/examples/webrtc
 ```
 
 ### Open example page
 
 [jsfiddle.net](https://jsfiddle.net/gh/get/library/pure/pion/mediadevices/tree/master/examples/internal/jsfiddle/audio-and-video) you should see two text-areas and a 'Start Session' button
 
-### Run simple with your browsers SessionDescription as stdin
+### Run webrtc with your browsers SessionDescription as stdin
 
 In the jsfiddle the top textarea is your browser, copy that and:
 
 #### Linux
 
-Run `echo $BROWSER_SDP | simple`
+Run `echo $BROWSER_SDP | webrtc`
 
-### Input simple's SessionDescription into your browser
+### Input webrtc's SessionDescription into your browser
 
-Copy the text that `simple` just emitted and copy into second text area
+Copy the text that `webrtc` just emitted and copy into second text area
 
 ### Hit 'Start Session' in jsfiddle, enjoy your video!
 

examples/webrtc/main.go

@@ -5,19 +5,18 @@ import (
 
 	"github.com/pion/mediadevices"
 	"github.com/pion/mediadevices/examples/internal/signal"
-	"github.com/pion/mediadevices/pkg/codec"
 	"github.com/pion/mediadevices/pkg/frame"
 	"github.com/pion/mediadevices/pkg/prop"
 	"github.com/pion/webrtc/v2"
 
-	// This is required to use opus audio encoder
-	"github.com/pion/mediadevices/pkg/codec/opus"
-
-	// If you don't like vpx, you can also use x264 by importing as below
-	// "github.com/pion/mediadevices/pkg/codec/x264" // This is required to use h264 video encoder
+	// If you don't like x264, you can also use vpx by importing as below
+	// "github.com/pion/mediadevices/pkg/codec/vpx" // This is required to use VP8/VP9 video encoder
 	// or you can also use openh264 for alternative h264 implementation
 	// "github.com/pion/mediadevices/pkg/codec/openh264"
-	"github.com/pion/mediadevices/pkg/codec/vpx" // This is required to use VP8/VP9 video encoder
+	// or if you use a raspberry pi like, you can use mmal for using its hardware encoder
+	// "github.com/pion/mediadevices/pkg/codec/mmal"
+	"github.com/pion/mediadevices/pkg/codec/opus" // This is required to use opus audio encoder
+	"github.com/pion/mediadevices/pkg/codec/x264" // This is required to use h264 video encoder
 
 	// Note: If you don't have a camera or microphone or your adapters are not supported,
 	//       you can always swap your adapters with our dummy adapters below.
@@ -45,7 +44,23 @@ func main() {
 	signal.Decode(signal.MustReadStdin(), &offer)
 
 	// Create a new RTCPeerConnection
+	x264Params, err := x264.NewParams()
+	if err != nil {
+		panic(err)
+	}
+	x264Params.BitRate = 500_000 // 500kbps
+
+	opusParams, err := opus.NewParams()
+	if err != nil {
+		panic(err)
+	}
+	codecSelector := mediadevices.NewCodecSelector(
+		mediadevices.WithVideoEncoders(&x264Params),
+		mediadevices.WithAudioEncoders(&opusParams),
+	)
+
 	mediaEngine := webrtc.MediaEngine{}
+	codecSelector.Populate(&mediaEngine)
 	if err := mediaEngine.PopulateFromSDP(offer); err != nil {
 		panic(err)
 	}
@@ -61,44 +76,33 @@ func main() {
 		fmt.Printf("Connection State has changed %s \n", connectionState.String())
 	})
 
-	md := mediadevices.NewMediaDevices(peerConnection)
-
-	opusParams, err := opus.NewParams()
-	if err != nil {
-		panic(err)
-	}
-	opusParams.BitRate = 32000 // 32kbps
-
-	vp8Params, err := vpx.NewVP8Params()
-	if err != nil {
-		panic(err)
-	}
-	vp8Params.BitRate = 100000 // 100kbps
-
-	s, err := md.GetUserMedia(mediadevices.MediaStreamConstraints{
-		Audio: func(c *mediadevices.MediaTrackConstraints) {
-			c.Enabled = true
-			c.AudioEncoderBuilders = []codec.AudioEncoderBuilder{&opusParams}
-		},
+	s, err := mediadevices.GetUserMedia(mediadevices.MediaStreamConstraints{
 		Video: func(c *mediadevices.MediaTrackConstraints) {
 			c.FrameFormat = prop.FrameFormat(frame.FormatYUY2)
-			c.Enabled = true
 			c.Width = prop.Int(640)
 			c.Height = prop.Int(480)
-			c.VideoEncoderBuilders = []codec.VideoEncoderBuilder{&vp8Params}
 		},
+		Audio: func(c *mediadevices.MediaTrackConstraints) {
+		},
+		Codec: codecSelector,
 	})
 	if err != nil {
 		panic(err)
 	}
 
 	for _, tracker := range s.GetTracks() {
-		t := tracker.Track()
 		tracker.OnEnded(func(err error) {
-			fmt.Printf("Track (ID: %s, Label: %s) ended with error: %v\n",
-				t.ID(), t.Label(), err)
+			fmt.Printf("Track (ID: %s) ended with error: %v\n",
+				tracker.ID(), err)
 		})
-		_, err = peerConnection.AddTransceiverFromTrack(t,
+
+		// In Pion/webrtc v3, bind will be called automatically after SDP negotiation
+		webrtcTrack, err := tracker.Bind(peerConnection)
+		if err != nil {
+			panic(err)
+		}
+
+		_, err = peerConnection.AddTransceiverFromTrack(webrtcTrack,
 			webrtc.RtpTransceiverInit{
 				Direction: webrtc.RTPTransceiverDirectionSendonly,
 			},

go.mod

@@ -4,10 +4,11 @@ go 1.13
 
 require (
 	github.com/blackjack/webcam v0.0.0-20200313125108-10ed912a8539
-	github.com/jfreymuth/pulse v0.0.0-20200817093420-a82ccdb5e8aa
-	github.com/lherman-cs/opus v0.0.0-20200223204610-6a4b98199ea4
+	github.com/jfreymuth/pulse v0.0.0-20201014123913-1e525c426c93
+	github.com/lherman-cs/opus v0.0.2
+	github.com/pion/logging v0.2.2
 	github.com/pion/webrtc/v2 v2.2.26
 	github.com/satori/go.uuid v1.2.0
-	golang.org/x/image v0.0.0-20200801110659-972c09e46d76
-	golang.org/x/sys v0.0.0-20200831180312-196b9ba8737a
+	golang.org/x/image v0.0.0-20200927104501-e162460cd6b5
+	golang.org/x/sys v0.0.0-20200930185726-fdedc70b468f
 )

go.sum

@@ -15,15 +15,15 @@ github.com/google/uuid v1.1.1 h1:Gkbcsh/GbpXz7lPftLA3P6TYMwjCLYm83jiFQZF/3gY=
 github.com/google/uuid v1.1.1/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
 github.com/hpcloud/tail v1.0.0 h1:nfCOvKYfkgYP8hkirhJocXT2+zOD8yUNjXaWfTlyFKI=
 github.com/hpcloud/tail v1.0.0/go.mod h1:ab1qPbhIpdTxEkNHXyeSf5vhxWSCs/tWer42PpOxQnU=
-github.com/jfreymuth/pulse v0.0.0-20200817093420-a82ccdb5e8aa h1:qUZIj5+D3UDgfshNe8Cz/9maOxe8ddt43qwQH9vEEC8=
-github.com/jfreymuth/pulse v0.0.0-20200817093420-a82ccdb5e8aa/go.mod h1:cpYspI6YljhkUf1WLXLLDmeaaPFc3CnGLjDZf9dZ4no=
+github.com/jfreymuth/pulse v0.0.0-20201014123913-1e525c426c93 h1:gDcaH96SZ7q1JU6hj0tSv8FiuqadFERU17lLxhphLa8=
+github.com/jfreymuth/pulse v0.0.0-20201014123913-1e525c426c93/go.mod h1:cpYspI6YljhkUf1WLXLLDmeaaPFc3CnGLjDZf9dZ4no=
 github.com/kr/pretty v0.1.0 h1:L/CwN0zerZDmRFUapSPitk6f+Q3+0za1rQkzVuMiMFI=
 github.com/kr/pretty v0.1.0/go.mod h1:dAy3ld7l9f0ibDNOQOHHMYYIIbhfbHSm3C4ZsoJORNo=
 github.com/kr/pty v1.1.1/go.mod h1:pFQYn66WHrOpPYNljwOMqo10TkYh1fy3cYio2l3bCsQ=
 github.com/kr/text v0.1.0 h1:45sCR5RtlFHMR4UwH9sdQ5TC8v0qDQCHnXt+kaKSTVE=
 github.com/kr/text v0.1.0/go.mod h1:4Jbv+DJW3UT/LiOwJeYQe1efqtUx/iVham/4vfdArNI=
-github.com/lherman-cs/opus v0.0.0-20200223204610-6a4b98199ea4 h1:2ydMA2KbxRkYmIw3R8Me8dn90bejxBR4MKYXJ5THK3I=
-github.com/lherman-cs/opus v0.0.0-20200223204610-6a4b98199ea4/go.mod h1:v9KQvlDYMuvlwniumBVMlrB0VHQvyTgxNvaXjPmTmps=
+github.com/lherman-cs/opus v0.0.2 h1:fE9Du3NKXDBztqvoTd6P2y9eJ9vgIHahGK8yQostnhA=
+github.com/lherman-cs/opus v0.0.2/go.mod h1:v9KQvlDYMuvlwniumBVMlrB0VHQvyTgxNvaXjPmTmps=
 github.com/lucas-clemente/quic-go v0.7.1-0.20190401152353-907071221cf9 h1:tbuodUh2vuhOVZAdW3NEUvosFHUMJwUNl7jk/VSEiwc=
 github.com/lucas-clemente/quic-go v0.7.1-0.20190401152353-907071221cf9/go.mod h1:PpMmPfPKO9nKJ/psF49ESTAGQSdfXxlg1otPbEB2nOw=
 github.com/marten-seemann/qtls v0.2.3 h1:0yWJ43C62LsZt08vuQJDK1uC1czUc3FJeCLPoNAI4vA=
@@ -35,7 +35,6 @@ github.com/onsi/gomega v1.4.3 h1:RE1xgDvH7imwFD45h+u2SgIfERHlS2yNG4DObb5BSKU=
 github.com/onsi/gomega v1.4.3/go.mod h1:ex+gbHU/CVuBBDIJjb2X0qEXbFg53c61hWP/1CpauHY=
 github.com/pion/datachannel v1.4.21 h1:3ZvhNyfmxsAqltQrApLPQMhSFNA+aT87RqyCq4OXmf0=
 github.com/pion/datachannel v1.4.21/go.mod h1:oiNyP4gHx2DIwRzX/MFyH0Rz/Gz05OgBlayAI2hAWjg=
-github.com/pion/dtls/v2 v2.0.1 h1:ddE7+V0faYRbyh4uPsRZ2vLdRrjVZn+wmCfI7jlBfaA=
 github.com/pion/dtls/v2 v2.0.1/go.mod h1:uMQkz2W0cSqY00xav7WByQ4Hb+18xeQh2oH2fRezr5U=
 github.com/pion/dtls/v2 v2.0.2 h1:FHCHTiM182Y8e15aFTiORroiATUI16ryHiQh8AIOJ1E=
 github.com/pion/dtls/v2 v2.0.2/go.mod h1:27PEO3MDdaCfo21heT59/vsdmZc0zMt9wQPcSlLu/1I=
@@ -63,9 +62,7 @@ github.com/pion/srtp v1.5.1/go.mod h1:B+QgX5xPeQTNc1CJStJPHzOlHK66ViMDWTT0HZTCkc
 github.com/pion/stun v0.3.5 h1:uLUCBCkQby4S1cf6CGuR9QrVOKcvUwFeemaC865QHDg=
 github.com/pion/stun v0.3.5/go.mod h1:gDMim+47EeEtfWogA37n6qXZS88L5V6LqFcf+DZA2UA=
 github.com/pion/transport v0.6.0/go.mod h1:iWZ07doqOosSLMhZ+FXUTq+TamDoXSllxpbGcfkCmbE=
-github.com/pion/transport v0.8.10 h1:lTiobMEw2PG6BH/mgIVqTV2mBp/mPT+IJLaN8ZxgdHk=
 github.com/pion/transport v0.8.10/go.mod h1:tBmha/UCjpum5hqTWhfAEs3CO4/tHSg0MYRhSzR+CZ8=
-github.com/pion/transport v0.10.0 h1:9M12BSneJm6ggGhJyWpDveFOstJsTiQjkLf4M44rm80=
 github.com/pion/transport v0.10.0/go.mod h1:BnHnUipd0rZQyTVB2SBGojFHT9CBt5C5TcsJSQGkvSE=
 github.com/pion/transport v0.10.1 h1:2W+yJT+0mOQ160ThZYUx5Zp2skzshiNgxrNE9GUfhJM=
 github.com/pion/transport v0.10.1/go.mod h1:PBis1stIILMiis0PewDw91WJeLJkyIMcEk+DwKOzf4A=
@@ -84,29 +81,23 @@ github.com/satori/go.uuid v1.2.0/go.mod h1:dA0hQrYB0VpLJoorglMZABFdXlWrHn1NEOzdh
 github.com/sclevine/agouti v3.0.0+incompatible/go.mod h1:b4WX9W9L1sfQKXeJf1mUTLZKJ48R1S7H23Ji7oFO5Bw=
 github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
 github.com/stretchr/testify v1.3.0/go.mod h1:M5WIy9Dh21IEIfnGCwXGc5bZfKNJtfHm1UVUgZn+9EI=
-github.com/stretchr/testify v1.4.0 h1:2E4SXV/wtOkTonXsotYi4li6zVWxYlZuYNCXe9XRJyk=
 github.com/stretchr/testify v1.4.0/go.mod h1:j7eGeouHqKxXV5pUuKE4zz7dFj8WfuZ+81PSLYec5m4=
-github.com/stretchr/testify v1.5.1 h1:nOGnQDM7FYENwehXlg/kFVnos3rEvtKTjRvOWSzb6H4=
 github.com/stretchr/testify v1.5.1/go.mod h1:5W2xD1RspED5o8YsWQXVCued0rvSQ+mT+I5cxcmMvtA=
 github.com/stretchr/testify v1.6.1 h1:hDPOHmpOpP40lSULcqw7IrRb/u7w6RpDC9399XyoNd0=
 github.com/stretchr/testify v1.6.1/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
 golang.org/x/crypto v0.0.0-20190228161510-8dd112bcdc25/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=
 golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=
-golang.org/x/crypto v0.0.0-20200602180216-279210d13fed h1:g4KENRiCMEx58Q7/ecwfT0N2o8z35Fnbsjig/Alf2T4=
 golang.org/x/crypto v0.0.0-20200602180216-279210d13fed/go.mod h1:LzIPMQfyMNhhGPhUkYOs5KpL4U8rLKemX1yGLhDgUto=
-golang.org/x/crypto v0.0.0-20200622213623-75b288015ac9 h1:psW17arqaxU48Z5kZ0CQnkZWQJsqcURM6tKiBApRjXI=
 golang.org/x/crypto v0.0.0-20200622213623-75b288015ac9/go.mod h1:LzIPMQfyMNhhGPhUkYOs5KpL4U8rLKemX1yGLhDgUto=
 golang.org/x/crypto v0.0.0-20200709230013-948cd5f35899 h1:DZhuSZLsGlFL4CmhA8BcRA0mnthyA/nZ00AqCUo7vHg=
 golang.org/x/crypto v0.0.0-20200709230013-948cd5f35899/go.mod h1:LzIPMQfyMNhhGPhUkYOs5KpL4U8rLKemX1yGLhDgUto=
-golang.org/x/image v0.0.0-20200801110659-972c09e46d76 h1:U7GPaoQyQmX+CBRWXKrvRzWTbd+slqeSh8uARsIyhAw=
-golang.org/x/image v0.0.0-20200801110659-972c09e46d76/go.mod h1:FeLwcggjj3mMvU+oOTbSwawSJRM1uh48EjtB4UJZlP0=
+golang.org/x/image v0.0.0-20200927104501-e162460cd6b5 h1:QelT11PB4FXiDEXucrfNckHoFxwt8USGY1ajP1ZF5lM=
+golang.org/x/image v0.0.0-20200927104501-e162460cd6b5/go.mod h1:FeLwcggjj3mMvU+oOTbSwawSJRM1uh48EjtB4UJZlP0=
 golang.org/x/net v0.0.0-20180906233101-161cd47e91fd/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
 golang.org/x/net v0.0.0-20190404232315-eb5bcb51f2a3/go.mod h1:t9HGtf8HONx5eT2rtn7q6eTqICYqUVnKs3thJo3Qplg=
 golang.org/x/net v0.0.0-20191126235420-ef20fe5d7933/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
 golang.org/x/net v0.0.0-20200226121028-0de0cce0169b/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
-golang.org/x/net v0.0.0-20200602114024-627f9648deb9 h1:pNX+40auqi2JqRfOP1akLGtYcn15TUbkhwuCO3foqqM=
 golang.org/x/net v0.0.0-20200602114024-627f9648deb9/go.mod h1:qpuaurCH72eLCgpAm/N6yyVIVM9cpaDIP3A8BGJEC5A=
-golang.org/x/net v0.0.0-20200625001655-4c5254603344 h1:vGXIOMxbNfDTk/aXCmfdLgkrSV+Z2tcbze+pEc3v5W4=
 golang.org/x/net v0.0.0-20200625001655-4c5254603344/go.mod h1:/O7V0waA8r7cgGh81Ro3o1hOxt32SMVPicZroKQ2sZA=
 golang.org/x/net v0.0.0-20200707034311-ab3426394381 h1:VXak5I6aEWmAXeQjA+QSZzlgNrpq9mjcfDemuexIKsU=
 golang.org/x/net v0.0.0-20200707034311-ab3426394381/go.mod h1:/O7V0waA8r7cgGh81Ro3o1hOxt32SMVPicZroKQ2sZA=
@@ -117,10 +108,9 @@ golang.org/x/sys v0.0.0-20190215142949-d0b11bdaac8a/go.mod h1:STP8DvDyc/dI5b8T5h
 golang.org/x/sys v0.0.0-20190228124157-a34e9553db1e/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20190412213103-97732733099d/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20200323222414-85ca7c5b95cd/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
-golang.org/x/sys v0.0.0-20200724161237-0e2f3a69832c h1:UIcGWL6/wpCfyGuJnRFJRurA+yj8RrW7Q6x2YMCXt6c=
 golang.org/x/sys v0.0.0-20200724161237-0e2f3a69832c/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
-golang.org/x/sys v0.0.0-20200831180312-196b9ba8737a h1:i47hUS795cOydZI4AwJQCKXOr4BvxzvikwDoDtHhP2Y=
-golang.org/x/sys v0.0.0-20200831180312-196b9ba8737a/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
+golang.org/x/sys v0.0.0-20200930185726-fdedc70b468f h1:+Nyd8tzPX9R7BWHguqsrbFdRx3WQ/1ib8I44HXV5yTA=
+golang.org/x/sys v0.0.0-20200930185726-fdedc70b468f/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/text v0.3.0 h1:g61tztE5qeGQ89tm6NTjjM9VPIm088od1l6aSorWRWg=
 golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
 golang.org/x/xerrors v0.0.0-20191204190536-9bdfabe68543 h1:E7g+9GITq07hpfrRu66IVDexMakfv52eLZ2CXBWiKr4=

internal/logging/logging.go

@@ -0,0 +1,11 @@
+package logging
+
+import (
+	"github.com/pion/logging"
+)
+
+var loggerFactory = logging.NewDefaultLoggerFactory()
+
+func NewLogger(scope string) logging.LeveledLogger {
+	return loggerFactory.NewLogger(scope)
+}

logging.go

@@ -0,0 +1,7 @@
+package mediadevices
+
+import (
+	"github.com/pion/mediadevices/internal/logging"
+)
+
+var logger = logging.NewLogger("mediadevices")

mediadeviceinfo.go

@@ -7,7 +7,7 @@ type MediaDeviceType int
 
 // MediaDeviceType definitions.
 const (
-	VideoInput MediaDeviceType = iota
+	VideoInput MediaDeviceType = iota + 1
 	AudioInput
 	AudioOutput
 )

mediadevices.go

@@ -7,95 +7,26 @@ import (
 
 	"github.com/pion/mediadevices/pkg/driver"
 	"github.com/pion/mediadevices/pkg/prop"
-	"github.com/pion/webrtc/v2"
 )
 
 var errNotFound = fmt.Errorf("failed to find the best driver that fits the constraints")
 
-// MediaDevices is an interface that's defined on https://developer.mozilla.org/en-US/docs/Web/API/MediaDevices
-type MediaDevices interface {
-	GetDisplayMedia(constraints MediaStreamConstraints) (MediaStream, error)
-	GetUserMedia(constraints MediaStreamConstraints) (MediaStream, error)
-	EnumerateDevices() []MediaDeviceInfo
-}
-
-// NewMediaDevices creates MediaDevices interface that provides access to connected media input devices
-// like cameras and microphones, as well as screen sharing.
-// In essence, it lets you obtain access to any hardware source of media data.
-func NewMediaDevices(pc *webrtc.PeerConnection, opts ...MediaDevicesOption) MediaDevices {
-	codecs := make(map[webrtc.RTPCodecType][]*webrtc.RTPCodec)
-	for _, kind := range []webrtc.RTPCodecType{
-		webrtc.RTPCodecTypeAudio,
-		webrtc.RTPCodecTypeVideo,
-	} {
-		codecs[kind] = pc.GetRegisteredRTPCodecs(kind)
-	}
-	return NewMediaDevicesFromCodecs(codecs, opts...)
-}
-
-// NewMediaDevicesFromCodecs creates MediaDevices interface from lists of the available codecs
-// that provides access to connected media input devices like cameras and microphones,
-// as well as screen sharing.
-// In essence, it lets you obtain access to any hardware source of media data.
-func NewMediaDevicesFromCodecs(codecs map[webrtc.RTPCodecType][]*webrtc.RTPCodec, opts ...MediaDevicesOption) MediaDevices {
-	mdo := MediaDevicesOptions{
-		codecs:         codecs,
-		trackGenerator: defaultTrackGenerator,
-	}
-	for _, o := range opts {
-		o(&mdo)
-	}
-	return &mediaDevices{
-		MediaDevicesOptions: mdo,
-	}
-}
-
-// TrackGenerator is a function to create new track.
-type TrackGenerator func(payloadType uint8, ssrc uint32, id, label string, codec *webrtc.RTPCodec) (LocalTrack, error)
-
-var defaultTrackGenerator = TrackGenerator(func(pt uint8, ssrc uint32, id, label string, codec *webrtc.RTPCodec) (LocalTrack, error) {
-	return webrtc.NewTrack(pt, ssrc, id, label, codec)
-})
-
-type mediaDevices struct {
-	MediaDevicesOptions
-}
-
-// MediaDevicesOptions stores parameters used by MediaDevices.
-type MediaDevicesOptions struct {
-	codecs         map[webrtc.RTPCodecType][]*webrtc.RTPCodec
-	trackGenerator TrackGenerator
-}
-
-// MediaDevicesOption is a type of MediaDevices functional option.
-type MediaDevicesOption func(*MediaDevicesOptions)
-
-// WithTrackGenerator specifies a TrackGenerator to use customized track.
-func WithTrackGenerator(gen TrackGenerator) MediaDevicesOption {
-	return func(o *MediaDevicesOptions) {
-		o.trackGenerator = gen
-	}
-}
-
 // GetDisplayMedia prompts the user to select and grant permission to capture the contents
 // of a display or portion thereof (such as a window) as a MediaStream.
 // Reference: https://developer.mozilla.org/en-US/docs/Web/API/MediaDevices/getDisplayMedia
-func (m *mediaDevices) GetDisplayMedia(constraints MediaStreamConstraints) (MediaStream, error) {
-	trackers := make([]Tracker, 0)
+func GetDisplayMedia(constraints MediaStreamConstraints) (MediaStream, error) {
+	trackers := make([]Track, 0)
 
 	cleanTrackers := func() {
 		for _, t := range trackers {
-			t.Stop()
+			t.Close()
 		}
 	}
 
 	var videoConstraints MediaTrackConstraints
 	if constraints.Video != nil {
 		constraints.Video(&videoConstraints)
-	}
-
-	if videoConstraints.Enabled {
-		tracker, err := m.selectScreen(videoConstraints)
+		tracker, err := selectScreen(videoConstraints, constraints.Codec)
 		if err != nil {
 			cleanTrackers()
 			return nil, err
@@ -116,27 +47,20 @@ func (m *mediaDevices) GetDisplayMedia(constraints MediaStreamConstraints) (Medi
 // GetUserMedia prompts the user for permission to use a media input which produces a MediaStream
 // with tracks containing the requested types of media.
 // Reference: https://developer.mozilla.org/en-US/docs/Web/API/MediaDevices/getUserMedia
-func (m *mediaDevices) GetUserMedia(constraints MediaStreamConstraints) (MediaStream, error) {
+func GetUserMedia(constraints MediaStreamConstraints) (MediaStream, error) {
 	// TODO: It should return media stream based on constraints
-	trackers := make([]Tracker, 0)
+	trackers := make([]Track, 0)
 
 	cleanTrackers := func() {
 		for _, t := range trackers {
-			t.Stop()
+			t.Close()
 		}
 	}
 
 	var videoConstraints, audioConstraints MediaTrackConstraints
 	if constraints.Video != nil {
 		constraints.Video(&videoConstraints)
-	}
-
-	if constraints.Audio != nil {
-		constraints.Audio(&audioConstraints)
-	}
-
-	if videoConstraints.Enabled {
-		tracker, err := m.selectVideo(videoConstraints)
+		tracker, err := selectVideo(videoConstraints, constraints.Codec)
 		if err != nil {
 			cleanTrackers()
 			return nil, err
@@ -145,8 +69,9 @@ func (m *mediaDevices) GetUserMedia(constraints MediaStreamConstraints) (MediaSt
 		trackers = append(trackers, tracker)
 	}
 
-	if audioConstraints.Enabled {
-		tracker, err := m.selectAudio(audioConstraints)
+	if constraints.Audio != nil {
+		constraints.Audio(&audioConstraints)
+		tracker, err := selectAudio(audioConstraints, constraints.Codec)
 		if err != nil {
 			cleanTrackers()
 			return nil, err
@@ -195,12 +120,15 @@ func queryDriverProperties(filter driver.FilterFn) map[driver.Driver][]prop.Medi
 func selectBestDriver(filter driver.FilterFn, constraints MediaTrackConstraints) (driver.Driver, MediaTrackConstraints, error) {
 	var bestDriver driver.Driver
 	var bestProp prop.Media
+	var foundPropertiesLog []string
 	minFitnessDist := math.Inf(1)
 
+	foundPropertiesLog = append(foundPropertiesLog, "\n============ Found Properties ============")
 	driverProperties := queryDriverProperties(filter)
 	for d, props := range driverProperties {
 		priority := float64(d.Info().Priority)
 		for _, p := range props {
+			foundPropertiesLog = append(foundPropertiesLog, p.String())
 			fitnessDist, ok := constraints.MediaConstraints.FitnessDistance(p)
 			if !ok {
 				continue
@@ -214,33 +142,25 @@ func selectBestDriver(filter driver.FilterFn, constraints MediaTrackConstraints)
 		}
 	}
 
+	foundPropertiesLog = append(foundPropertiesLog, "=============== Constraints ==============")
+	foundPropertiesLog = append(foundPropertiesLog, constraints.String())
+	foundPropertiesLog = append(foundPropertiesLog, "================ Best Fit ================")
+
 	if bestDriver == nil {
-		var foundProperties []string
-		for _, props := range driverProperties {
-			for _, p := range props {
-				foundProperties = append(foundProperties, fmt.Sprint(&p))
-			}
-		}
-
-		err := fmt.Errorf(`%w:
-============ Found Properties ============
-
-%s
-
-=============== Constraints ==============
-
-%s
-`, errNotFound, strings.Join(foundProperties, "\n\n"), &constraints)
-		return nil, MediaTrackConstraints{}, err
+		foundPropertiesLog = append(foundPropertiesLog, "Not found")
+		logger.Debug(strings.Join(foundPropertiesLog, "\n\n"))
+		return nil, MediaTrackConstraints{}, errNotFound
 	}
 
+	foundPropertiesLog = append(foundPropertiesLog, bestProp.String())
+	logger.Debug(strings.Join(foundPropertiesLog, "\n\n"))
 	constraints.selectedMedia = prop.Media{}
 	constraints.selectedMedia.MergeConstraints(constraints.MediaConstraints)
 	constraints.selectedMedia.Merge(bestProp)
 	return bestDriver, constraints, nil
 }
 
-func (m *mediaDevices) selectAudio(constraints MediaTrackConstraints) (Tracker, error) {
+func selectAudio(constraints MediaTrackConstraints, selector *CodecSelector) (Track, error) {
 	typeFilter := driver.FilterAudioRecorder()
 
 	d, c, err := selectBestDriver(typeFilter, constraints)
@@ -248,9 +168,9 @@ func (m *mediaDevices) selectAudio(constraints MediaTrackConstraints) (Tracker,
 		return nil, err
 	}
 
-	return newTrack(&m.MediaDevicesOptions, d, c)
+	return newTrackFromDriver(d, c, selector)
 }
-func (m *mediaDevices) selectVideo(constraints MediaTrackConstraints) (Tracker, error) {
+func selectVideo(constraints MediaTrackConstraints, selector *CodecSelector) (Track, error) {
 	typeFilter := driver.FilterVideoRecorder()
 	notScreenFilter := driver.FilterNot(driver.FilterDeviceType(driver.Screen))
 	filter := driver.FilterAnd(typeFilter, notScreenFilter)
@@ -260,10 +180,10 @@ func (m *mediaDevices) selectVideo(constraints MediaTrackConstraints) (Tracker,
 		return nil, err
 	}
 
-	return newTrack(&m.MediaDevicesOptions, d, c)
+	return newTrackFromDriver(d, c, selector)
 }
 
-func (m *mediaDevices) selectScreen(constraints MediaTrackConstraints) (Tracker, error) {
+func selectScreen(constraints MediaTrackConstraints, selector *CodecSelector) (Track, error) {
 	typeFilter := driver.FilterVideoRecorder()
 	screenFilter := driver.FilterDeviceType(driver.Screen)
 	filter := driver.FilterAnd(typeFilter, screenFilter)
@@ -273,10 +193,10 @@ func (m *mediaDevices) selectScreen(constraints MediaTrackConstraints) (Tracker,
 		return nil, err
 	}
 
-	return newTrack(&m.MediaDevicesOptions, d, c)
+	return newTrackFromDriver(d, c, selector)
 }
 
-func (m *mediaDevices) EnumerateDevices() []MediaDeviceInfo {
+func EnumerateDevices() []MediaDeviceInfo {
 	drivers := driver.GetManager().Query(
 		driver.FilterFn(func(driver.Driver) bool { return true }))
 	info := make([]MediaDeviceInfo, 0, len(drivers))

mediadevices_bench_test.go

@@ -0,0 +1,82 @@
+// +build e2e
+
+package mediadevices
+
+import (
+	"image"
+	"sync"
+	"testing"
+
+	"github.com/pion/mediadevices/pkg/codec/x264"
+	"github.com/pion/mediadevices/pkg/frame"
+)
+
+type mockVideoSource struct {
+	width, height int
+	pool          sync.Pool
+	decoder       frame.Decoder
+}
+
+func newMockVideoSource(width, height int) *mockVideoSource {
+	decoder, err := frame.NewDecoder(frame.FormatYUY2)
+	if err != nil {
+		panic(err)
+	}
+
+	return &mockVideoSource{
+		width:  width,
+		height: height,
+		pool: sync.Pool{
+			New: func() interface{} {
+				resolution := width * height
+				return make([]byte, resolution*2)
+			},
+		},
+		decoder: decoder,
+	}
+}
+
+func (source *mockVideoSource) ID() string   { return "" }
+func (source *mockVideoSource) Close() error { return nil }
+func (source *mockVideoSource) Read() (image.Image, func(), error) {
+	raw := source.pool.Get().([]byte)
+	decoded, release, err := source.decoder.Decode(raw, source.width, source.height)
+	source.pool.Put(raw)
+	if err != nil {
+		return nil, nil, err
+	}
+
+	return decoded, release, nil
+}
+
+func BenchmarkEndToEnd(b *testing.B) {
+	params, err := x264.NewParams()
+	if err != nil {
+		b.Fatal(err)
+	}
+	params.BitRate = 300_000
+
+	videoSource := newMockVideoSource(1920, 1080)
+	track := NewVideoTrack(videoSource, nil).(*VideoTrack)
+	defer track.Close()
+
+	reader := track.NewReader(false)
+	inputProp, err := detectCurrentVideoProp(track.Broadcaster)
+	if err != nil {
+		b.Fatal(err)
+	}
+
+	encodedReader, err := params.BuildVideoEncoder(reader, inputProp)
+	if err != nil {
+		b.Fatal(err)
+	}
+	defer encodedReader.Close()
+
+	for i := 0; i < b.N; i++ {
+		_, release, err := encodedReader.Read()
+		if err != nil {
+			b.Fatal(err)
+		}
+		release()
+	}
+}

mediadevices_test.go

@@ -1,91 +1,42 @@
 package mediadevices
 
 import (
-	"errors"
 	"io"
 	"testing"
 	"time"
 
-	"github.com/pion/webrtc/v2"
-	"github.com/pion/webrtc/v2/pkg/media"
-
-	"github.com/pion/mediadevices/pkg/codec"
 	"github.com/pion/mediadevices/pkg/driver"
 	_ "github.com/pion/mediadevices/pkg/driver/audiotest"
 	_ "github.com/pion/mediadevices/pkg/driver/videotest"
-	"github.com/pion/mediadevices/pkg/io/audio"
-	"github.com/pion/mediadevices/pkg/io/video"
 	"github.com/pion/mediadevices/pkg/prop"
 )
 
 func TestGetUserMedia(t *testing.T) {
-	videoParams := mockParams{
-		BaseParams: codec.BaseParams{
-			BitRate: 100000,
-		},
-		name: "MockVideo",
-	}
-	audioParams := mockParams{
-		BaseParams: codec.BaseParams{
-			BitRate: 32000,
-		},
-		name: "MockAudio",
-	}
-	md := NewMediaDevicesFromCodecs(
-		map[webrtc.RTPCodecType][]*webrtc.RTPCodec{
-			webrtc.RTPCodecTypeVideo: {
-				{Type: webrtc.RTPCodecTypeVideo, Name: "MockVideo", PayloadType: 1},
-			},
-			webrtc.RTPCodecTypeAudio: {
-				{Type: webrtc.RTPCodecTypeAudio, Name: "MockAudio", PayloadType: 2},
-			},
-		},
-		WithTrackGenerator(
-			func(_ uint8, _ uint32, id, _ string, codec *webrtc.RTPCodec) (
-				LocalTrack, error,
-			) {
-				return newMockTrack(codec, id), nil
-			},
-		),
-	)
 	constraints := MediaStreamConstraints{
 		Video: func(c *MediaTrackConstraints) {
-			c.Enabled = true
 			c.Width = prop.Int(640)
 			c.Height = prop.Int(480)
-			params := videoParams
-			c.VideoEncoderBuilders = []codec.VideoEncoderBuilder{&params}
 		},
 		Audio: func(c *MediaTrackConstraints) {
-			c.Enabled = true
-			params := audioParams
-			c.AudioEncoderBuilders = []codec.AudioEncoderBuilder{&params}
 		},
 	}
 	constraintsWrong := MediaStreamConstraints{
 		Video: func(c *MediaTrackConstraints) {
-			c.Enabled = true
-			c.Width = prop.Int(640)
+			c.Width = prop.IntExact(10000)
 			c.Height = prop.Int(480)
-			params := videoParams
-			params.BitRate = 0
-			c.VideoEncoderBuilders = []codec.VideoEncoderBuilder{&params}
 		},
 		Audio: func(c *MediaTrackConstraints) {
-			c.Enabled = true
-			params := audioParams
-			c.AudioEncoderBuilders = []codec.AudioEncoderBuilder{&params}
 		},
 	}
 
 	// GetUserMedia with broken parameters
-	ms, err := md.GetUserMedia(constraintsWrong)
+	ms, err := GetUserMedia(constraintsWrong)
 	if err == nil {
 		t.Fatal("Expected error, but got nil")
 	}
 
 	// GetUserMedia with correct parameters
-	ms, err = md.GetUserMedia(constraints)
+	ms, err = GetUserMedia(constraints)
 	if err != nil {
 		t.Fatalf("Unexpected error: %v", err)
 	}
@@ -103,11 +54,11 @@ func TestGetUserMedia(t *testing.T) {
 	time.Sleep(50 * time.Millisecond)
 
 	for _, track := range tracks {
-		track.Stop()
+		track.Close()
 	}
 
 	// Stop and retry GetUserMedia
-	ms, err = md.GetUserMedia(constraints)
+	ms, err = GetUserMedia(constraints)
 	if err != nil {
 		t.Fatalf("Failed to GetUserMedia after the previsous tracks stopped: %v", err)
 	}
@@ -124,104 +75,10 @@ func TestGetUserMedia(t *testing.T) {
 	}
 	time.Sleep(50 * time.Millisecond)
 	for _, track := range tracks {
-		track.Stop()
+		track.Close()
 	}
 }
 
-type mockTrack struct {
-	codec *webrtc.RTPCodec
-	id    string
-}
-
-func newMockTrack(codec *webrtc.RTPCodec, id string) *mockTrack {
-	return &mockTrack{
-		codec: codec,
-		id:    id,
-	}
-}
-
-func (t *mockTrack) WriteSample(s media.Sample) error {
-	return nil
-}
-
-func (t *mockTrack) Codec() *webrtc.RTPCodec {
-	return t.codec
-}
-
-func (t *mockTrack) ID() string {
-	return t.id
-}
-
-func (t *mockTrack) Kind() webrtc.RTPCodecType {
-	return t.codec.Type
-}
-
-type mockParams struct {
-	codec.BaseParams
-	name string
-}
-
-func (params *mockParams) Name() string {
-	return params.name
-}
-
-func (params *mockParams) BuildVideoEncoder(r video.Reader, p prop.Media) (codec.ReadCloser, error) {
-	if params.BitRate == 0 {
-		// This is a dummy error to test the failure condition.
-		return nil, errors.New("wrong codec parameter")
-	}
-	return &mockVideoCodec{
-		r:      r,
-		closed: make(chan struct{}),
-	}, nil
-}
-
-func (params *mockParams) BuildAudioEncoder(r audio.Reader, p prop.Media) (codec.ReadCloser, error) {
-	return &mockAudioCodec{
-		r:      r,
-		closed: make(chan struct{}),
-	}, nil
-}
-
-type mockCodec struct{}
-
-func (e *mockCodec) SetBitRate(b int) error {
-	return nil
-}
-
-func (e *mockCodec) ForceKeyFrame() error {
-	return nil
-}
-
-type mockVideoCodec struct {
-	mockCodec
-	r      video.Reader
-	closed chan struct{}
-}
-
-func (m *mockVideoCodec) Read(b []byte) (int, error) {
-	if _, err := m.r.Read(); err != nil {
-		return 0, err
-	}
-	return len(b), nil
-}
-
-func (m *mockVideoCodec) Close() error { return nil }
-
-type mockAudioCodec struct {
-	mockCodec
-	r      audio.Reader
-	closed chan struct{}
-}
-
-func (m *mockAudioCodec) Read(b []byte) (int, error) {
-	if _, err := m.r.Read(); err != nil {
-		return 0, err
-	}
-	return len(b), nil
-}
-func (m *mockAudioCodec) Close() error { return nil }
-
 func TestSelectBestDriverConstraintsResultIsSetProperly(t *testing.T) {
 	filterFn := driver.FilterVideoRecorder()
 	drivers := driver.GetManager().Query(filterFn)

mediastream.go

@@ -2,89 +2,85 @@ package mediadevices
 
 import (
 	"sync"
-
-	"github.com/pion/webrtc/v2"
 )
 
 // MediaStream is an interface that represents a collection of existing tracks.
 type MediaStream interface {
 	// GetAudioTracks implements https://w3c.github.io/mediacapture-main/#dom-mediastream-getaudiotracks
-	GetAudioTracks() []Tracker
+	GetAudioTracks() []Track
 	// GetVideoTracks implements https://w3c.github.io/mediacapture-main/#dom-mediastream-getvideotracks
-	GetVideoTracks() []Tracker
+	GetVideoTracks() []Track
 	// GetTracks implements https://w3c.github.io/mediacapture-main/#dom-mediastream-gettracks
-	GetTracks() []Tracker
+	GetTracks() []Track
 	// AddTrack implements https://w3c.github.io/mediacapture-main/#dom-mediastream-addtrack
-	AddTrack(t Tracker)
+	AddTrack(t Track)
 	// RemoveTrack implements https://w3c.github.io/mediacapture-main/#dom-mediastream-removetrack
-	RemoveTrack(t Tracker)
+	RemoveTrack(t Track)
 }
 
 type mediaStream struct {
-	trackers map[string]Tracker
-	l        sync.RWMutex
+	tracks map[Track]struct{}
+	l      sync.RWMutex
 }
 
-const rtpCodecTypeDefault webrtc.RTPCodecType = 0
+const trackTypeDefault MediaDeviceType = 0
 
 // NewMediaStream creates a MediaStream interface that's defined in
 // https://w3c.github.io/mediacapture-main/#dom-mediastream
-func NewMediaStream(trackers ...Tracker) (MediaStream, error) {
-	m := mediaStream{trackers: make(map[string]Tracker)}
+func NewMediaStream(tracks ...Track) (MediaStream, error) {
+	m := mediaStream{tracks: make(map[Track]struct{})}
 
-	for _, tracker := range trackers {
-		id := tracker.LocalTrack().ID()
-		if _, ok := m.trackers[id]; !ok {
-			m.trackers[id] = tracker
+	for _, track := range tracks {
+		if _, ok := m.tracks[track]; !ok {
+			m.tracks[track] = struct{}{}
 		}
 	}
 
 	return &m, nil
 }
 
-func (m *mediaStream) GetAudioTracks() []Tracker {
-	return m.queryTracks(webrtc.RTPCodecTypeAudio)
+func (m *mediaStream) GetAudioTracks() []Track {
+	return m.queryTracks(AudioInput)
 }
 
-func (m *mediaStream) GetVideoTracks() []Tracker {
-	return m.queryTracks(webrtc.RTPCodecTypeVideo)
+func (m *mediaStream) GetVideoTracks() []Track {
+	return m.queryTracks(VideoInput)
 }
 
-func (m *mediaStream) GetTracks() []Tracker {
-	return m.queryTracks(rtpCodecTypeDefault)
+func (m *mediaStream) GetTracks() []Track {
+	return m.queryTracks(trackTypeDefault)
 }
 
 // queryTracks returns all tracks that are the same kind as t.
 // If t is 0, which is the default, queryTracks will return all the tracks.
-func (m *mediaStream) queryTracks(t webrtc.RTPCodecType) []Tracker {
+func (m *mediaStream) queryTracks(t MediaDeviceType) []Track {
 	m.l.RLock()
 	defer m.l.RUnlock()
 
-	result := make([]Tracker, 0)
-	for _, tracker := range m.trackers {
-		if tracker.LocalTrack().Kind() == t || t == rtpCodecTypeDefault {
-			result = append(result, tracker)
+	result := make([]Track, 0)
+	for track := range m.tracks {
+		if track.Kind() == t || t == trackTypeDefault {
+			result = append(result, track)
 		}
 	}
 
 	return result
 }
 
-func (m *mediaStream) AddTrack(t Tracker) {
+func (m *mediaStream) AddTrack(t Track) {
 	m.l.Lock()
 	defer m.l.Unlock()
 
-	id := t.LocalTrack().ID()
-	if _, ok := m.trackers[id]; ok {
+	if _, ok := m.tracks[t]; ok {
 		return
 	}
 
-	m.trackers[id] = t
+	m.tracks[t] = struct{}{}
 }
 
-func (m *mediaStream) RemoveTrack(t Tracker) {
+func (m *mediaStream) RemoveTrack(t Track) {
 	m.l.Lock()
 	defer m.l.Unlock()
 
-	delete(m.trackers, t.LocalTrack().ID())
+	delete(m.tracks, t)
 }

mediastream_test.go

@@ -0,0 +1,88 @@
+package mediadevices
+
+import (
+	"testing"
+
+	"github.com/pion/webrtc/v2"
+)
+
+type mockMediaStreamTrack struct {
+	kind MediaDeviceType
+}
+
+func (track *mockMediaStreamTrack) ID() string {
+	return ""
+}
+
+func (track *mockMediaStreamTrack) Close() error {
+	return nil
+}
+
+func (track *mockMediaStreamTrack) Kind() MediaDeviceType {
+	return track.kind
+}
+
+func (track *mockMediaStreamTrack) OnEnded(handler func(error)) {
+}
+
+func (track *mockMediaStreamTrack) Bind(pc *webrtc.PeerConnection) (*webrtc.Track, error) {
+	return nil, nil
+}
+
+func (track *mockMediaStreamTrack) Unbind(pc *webrtc.PeerConnection) error {
+	return nil
+}
+
+func TestMediaStreamFilters(t *testing.T) {
+	audioTracks := []Track{
+		&mockMediaStreamTrack{AudioInput},
+		&mockMediaStreamTrack{AudioInput},
+		&mockMediaStreamTrack{AudioInput},
+		&mockMediaStreamTrack{AudioInput},
+		&mockMediaStreamTrack{AudioInput},
+	}
+
+	videoTracks := []Track{
+		&mockMediaStreamTrack{VideoInput},
+		&mockMediaStreamTrack{VideoInput},
+		&mockMediaStreamTrack{VideoInput},
+	}
+
+	tracks := append(audioTracks, videoTracks...)
+	stream, err := NewMediaStream(tracks...)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	expect := func(t *testing.T, actual, expected []Track) {
+		if len(actual) != len(expected) {
+			t.Fatalf("%s: Expected to get %d trackers, but got %d trackers", t.Name(), len(expected), len(actual))
+		}
+
+		for _, a := range actual {
+			found := false
+			for _, e := range expected {
+				if e == a {
+					found = true
+					break
+				}
+			}
+
+			if !found {
+				t.Fatalf("%s: Expected to find %p in the query results", t.Name(), a)
+			}
+		}
+	}
+
+	t.Run("GetAudioTracks", func(t *testing.T) {
+		expect(t, stream.GetAudioTracks(), audioTracks)
+	})
+
+	t.Run("GetVideoTracks", func(t *testing.T) {
+		expect(t, stream.GetVideoTracks(), videoTracks)
+	})
+
+	t.Run("GetTracks", func(t *testing.T) {
+		expect(t, stream.GetTracks(), tracks)
+	})
+}

mediastreamconstraints.go

@@ -1,40 +1,18 @@
 package mediadevices
 
 import (
-	"github.com/pion/mediadevices/pkg/codec"
-	"github.com/pion/mediadevices/pkg/io/audio"
-	"github.com/pion/mediadevices/pkg/io/video"
 	"github.com/pion/mediadevices/pkg/prop"
 )
 
 type MediaStreamConstraints struct {
 	Audio MediaOption
 	Video MediaOption
+	Codec *CodecSelector
 }
 
 // MediaTrackConstraints represents https://w3c.github.io/mediacapture-main/#dom-mediatrackconstraints
 type MediaTrackConstraints struct {
 	prop.MediaConstraints
-	Enabled bool
-	// VideoEncoderBuilders are codec builders that are used for encoding the video
-	// and later being used for sending the appropriate RTP payload type.
-	//
-	// If one encoder builder fails to build the codec, the next builder will be used,
-	// repeating until a codec builds. If no builders build successfully, an error is returned.
-	VideoEncoderBuilders []codec.VideoEncoderBuilder
-	// AudioEncoderBuilders are codec builders that are used for encoding the audio
-	// and later being used for sending the appropriate RTP payload type.
-	//
-	// If one encoder builder fails to build the codec, the next builder will be used,
-	// repeating until a codec builds. If no builders build successfully, an error is returned.
-	AudioEncoderBuilders []codec.AudioEncoderBuilder
-	// VideoTransform will be used to transform the video that's coming from the driver.
-	// So, basically it'll look like following: driver -> VideoTransform -> codec
-	VideoTransform video.TransformFunc
-	// AudioTransform will be used to transform the audio that's coming from the driver.
-	// So, basically it'll look like following: driver -> AudioTransform -> code
-	AudioTransform audio.TransformFunc
-
 	selectedMedia prop.Media
 }
 

meta.go

@@ -0,0 +1,35 @@
+package mediadevices
+
+import (
+	"github.com/pion/mediadevices/pkg/io/audio"
+	"github.com/pion/mediadevices/pkg/io/video"
+	"github.com/pion/mediadevices/pkg/prop"
+)
+
+// detectCurrentVideoProp is a small helper to get current video property
+func detectCurrentVideoProp(broadcaster *video.Broadcaster) (prop.Media, error) {
+	var currentProp prop.Media
+
+	// Since broadcaster has a ring buffer internally, a new reader will either read the last
+	// buffered frame or a new frame from the source. This also implies that no frame will be lost
+	// in any case.
+	metaReader := broadcaster.NewReader(false)
+	metaReader = video.DetectChanges(0, func(p prop.Media) { currentProp = p })(metaReader)
+	_, _, err := metaReader.Read()
+
+	return currentProp, err
+}
+
+// detectCurrentAudioProp is a small helper to get current audio property
+func detectCurrentAudioProp(broadcaster *audio.Broadcaster) (prop.Media, error) {
+	var currentProp prop.Media
+
+	// Since broadcaster has a ring buffer internally, a new reader will either read the last
+	// buffered frame or a new frame from the source. This also implies that no frame will be lost
+	// in any case.
+	metaReader := broadcaster.NewReader(false)
+	metaReader = audio.DetectChanges(0, func(p prop.Media) { currentProp = p })(metaReader)
+	_, _, err := metaReader.Read()
+
+	return currentProp, err
+}

meta_test.go

@@ -0,0 +1,98 @@
+package mediadevices
+
+import (
+	"image"
+	"testing"
+
+	"github.com/pion/mediadevices/pkg/io/audio"
+	"github.com/pion/mediadevices/pkg/io/video"
+	"github.com/pion/mediadevices/pkg/wave"
+)
+
+func TestDetectCurrentVideoProp(t *testing.T) {
+	resolution := image.Rect(0, 0, 4, 4)
+	first := image.NewRGBA(resolution)
+	first.Pix[0] = 1
+	second := image.NewRGBA(resolution)
+	second.Pix[0] = 2
+
+	isFirst := true
+	source := video.ReaderFunc(func() (image.Image, func(), error) {
+		if isFirst {
+			isFirst = true
+			return first, func() {}, nil
+		} else {
+			return second, func() {}, nil
+		}
+	})
+
+	broadcaster := video.NewBroadcaster(source, nil)
+
+	currentProp, err := detectCurrentVideoProp(broadcaster)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if currentProp.Width != resolution.Dx() {
+		t.Fatalf("Expect the actual width to be %d, but got %d", currentProp.Width, resolution.Dx())
+	}
+
+	if currentProp.Height != resolution.Dy() {
+		t.Fatalf("Expect the actual height to be %d, but got %d", currentProp.Height, resolution.Dy())
+	}
+
+	reader := broadcaster.NewReader(false)
+	img, _, err := reader.Read()
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	rgba := img.(*image.RGBA)
+	if rgba.Pix[0] != 1 {
+		t.Fatal("Expect the frame after reading the current prop is not the first frame")
+	}
+}
+
+func TestDetectCurrentAudioProp(t *testing.T) {
+	info := wave.ChunkInfo{
+		Len:          4,
+		Channels:     2,
+		SamplingRate: 48000,
+	}
+	first := wave.NewInt16Interleaved(info)
+	first.Data[0] = 1
+	second := wave.NewInt16Interleaved(info)
+	second.Data[0] = 2
+
+	isFirst := true
+	source := audio.ReaderFunc(func() (wave.Audio, func(), error) {
+		if isFirst {
+			isFirst = true
+			return first, func() {}, nil
+		} else {
+			return second, func() {}, nil
+		}
+	})
+
+	broadcaster := audio.NewBroadcaster(source, nil)
+
+	currentProp, err := detectCurrentAudioProp(broadcaster)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if currentProp.ChannelCount != info.Channels {
+		t.Fatalf("Expect the actual channel count to be %d, but got %d", currentProp.ChannelCount, info.Channels)
+	}
+
+	reader := broadcaster.NewReader(false)
+	chunk, _, err := reader.Read()
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	realChunk := chunk.(*wave.Int16Interleaved)
+	if realChunk.Data[0] != 1 {
+		t.Fatal("Expect the chunk after reading the current prop is not the first chunk")
+	}
+}

pkg/avfoundation/avfoundation_darwin.go

@@ -110,12 +110,12 @@ func (rc *ReadCloser) dataCb(data []byte) {
 // Read reads raw data, the format is determined by the media type and property:
 //   - For video, each call will return a frame.
 //   - For audio, each call will return a chunk which its size configured by Latency
-func (rc *ReadCloser) Read() ([]byte, error) {
+func (rc *ReadCloser) Read() ([]byte, func(), error) {
 	data, ok := <-rc.dataChan
 	if !ok {
-		return nil, io.EOF
+		return nil, func() {}, io.EOF
 	}
-	return data, nil
+	return data, func() {}, nil
 }
 
 // Close closes the capturing session, and no data will flow anymore

pkg/codec/codec.go

@@ -1,21 +1,45 @@
 package codec
 
 import (
-	"io"
-
 	"github.com/pion/mediadevices/pkg/io/audio"
 	"github.com/pion/mediadevices/pkg/io/video"
 	"github.com/pion/mediadevices/pkg/prop"
+	"github.com/pion/webrtc/v2"
 )
 
+// RTPCodec wraps webrtc.RTPCodec. RTPCodec might extend webrtc.RTPCodec in the future.
+type RTPCodec struct {
+	*webrtc.RTPCodec
+}
+
+// NewRTPH264Codec is a helper to create an H264 codec
+func NewRTPH264Codec(clockrate uint32) *RTPCodec {
+	return &RTPCodec{webrtc.NewRTPH264Codec(webrtc.DefaultPayloadTypeH264, clockrate)}
+}
+
+// NewRTPVP8Codec is a helper to create an VP8 codec
+func NewRTPVP8Codec(clockrate uint32) *RTPCodec {
+	return &RTPCodec{webrtc.NewRTPVP8Codec(webrtc.DefaultPayloadTypeVP8, clockrate)}
+}
+
+// NewRTPVP9Codec is a helper to create an VP9 codec
+func NewRTPVP9Codec(clockrate uint32) *RTPCodec {
+	return &RTPCodec{webrtc.NewRTPVP9Codec(webrtc.DefaultPayloadTypeVP9, clockrate)}
+}
+
+// NewRTPOpusCodec is a helper to create an Opus codec
+func NewRTPOpusCodec(clockrate uint32) *RTPCodec {
+	return &RTPCodec{webrtc.NewRTPOpusCodec(webrtc.DefaultPayloadTypeOpus, clockrate)}
+}
+
 // AudioEncoderBuilder is the interface that wraps basic operations that are
 // necessary to build the audio encoder.
 //
 // This interface is for codec implementors to provide codec specific params,
 // but still giving generality for the users.
 type AudioEncoderBuilder interface {
-	// Name represents the codec name
-	Name() string
+	// RTPCodec represents the codec metadata
+	RTPCodec() *RTPCodec
 	// BuildAudioEncoder builds audio encoder by given media params and audio input
 	BuildAudioEncoder(r audio.Reader, p prop.Media) (ReadCloser, error)
 }
@@ -26,15 +50,16 @@ type AudioEncoderBuilder interface {
 // This interface is for codec implementors to provide codec specific params,
 // but still giving generality for the users.
 type VideoEncoderBuilder interface {
-	// Name represents the codec name
-	Name() string
+	// RTPCodec represents the codec metadata
+	RTPCodec() *RTPCodec
 	// BuildVideoEncoder builds video encoder by given media params and video input
 	BuildVideoEncoder(r video.Reader, p prop.Media) (ReadCloser, error)
 }
 
 // ReadCloser is an io.ReadCloser with methods for rate limiting: SetBitRate and ForceKeyFrame
 type ReadCloser interface {
-	io.ReadCloser
+	Read() (b []byte, release func(), err error)
+	Close() error
 	// SetBitRate sets current target bitrate, lower bitrate means smaller data will be transmitted
 	// but this also means that the quality will also be lower.
 	SetBitRate(int) error

pkg/codec/mmal/bridge.h

@@ -0,0 +1,196 @@
+#include <interface/mmal/mmal.h>
+#include <interface/mmal/util/mmal_default_components.h>
+#include <interface/mmal/util/mmal_util_params.h>
+#include <stdbool.h>
+#include <stdint.h>
+#include <stdlib.h>
+
+#define CHK(__status, __msg)                                                                                           \
+  do {                                                                                                                 \
+    status.code = __status;                                                                                            \
+    if (status.code != MMAL_SUCCESS) {                                                                                 \
+      status.msg = __msg;                                                                                              \
+      goto CleanUp;                                                                                                    \
+    }                                                                                                                  \
+  } while (0)
+
+typedef struct Status {
+  MMAL_STATUS_T code;
+  const char *msg;
+} Status;
+
+typedef struct Slice {
+  uint8_t *data;
+  int len;
+} Slice;
+
+typedef struct Params {
+  int width, height;
+  uint32_t bitrate;
+  uint32_t key_frame_interval;
+} Params;
+
+typedef struct Encoder {
+  MMAL_COMPONENT_T *component;
+  MMAL_PORT_T *port_in, *port_out;
+  MMAL_QUEUE_T *queue_out;
+  MMAL_POOL_T *pool_in, *pool_out;
+} Encoder;
+
+Status enc_new(Params, Encoder *);
+Status enc_encode(Encoder *, Slice y, Slice cb, Slice cr, MMAL_BUFFER_HEADER_T **);
+Status enc_close(Encoder *);
+
+static void encoder_in_cb(MMAL_PORT_T *port, MMAL_BUFFER_HEADER_T *buffer) { mmal_buffer_header_release(buffer); }
+
+static void encoder_out_cb(MMAL_PORT_T *port, MMAL_BUFFER_HEADER_T *buffer) {
+  MMAL_QUEUE_T *queue = (MMAL_QUEUE_T *)port->userdata;
+  mmal_queue_put(queue, buffer);
+}
+
+Status enc_new(Params params, Encoder *encoder) {
+  Status status = {0};
+  bool created = false;
+
+  memset(encoder, 0, sizeof(Encoder));
+
+  CHK(mmal_component_create(MMAL_COMPONENT_DEFAULT_VIDEO_ENCODER, &encoder->component),
+      "Failed to create video encoder component");
+  created = true;
+
+  encoder->port_in = encoder->component->input[0];
+  encoder->port_in->format->type = MMAL_ES_TYPE_VIDEO;
+  encoder->port_in->format->encoding = MMAL_ENCODING_I420;
+  encoder->port_in->format->es->video.width = params.width;
+  encoder->port_in->format->es->video.height = params.height;
+  encoder->port_in->format->es->video.par.num = 1;
+  encoder->port_in->format->es->video.par.den = 1;
+  encoder->port_in->format->es->video.crop.x = 0;
+  encoder->port_in->format->es->video.crop.y = 0;
+  encoder->port_in->format->es->video.crop.width = params.width;
+  encoder->port_in->format->es->video.crop.height = params.height;
+  CHK(mmal_port_format_commit(encoder->port_in), "Failed to commit input port format");
+
+  encoder->port_out = encoder->component->output[0];
+  encoder->port_out->format->type = MMAL_ES_TYPE_VIDEO;
+  encoder->port_out->format->encoding = MMAL_ENCODING_H264;
+  encoder->port_out->format->bitrate = params.bitrate;
+  CHK(mmal_port_format_commit(encoder->port_out), "Failed to commit output port format");
+
+  MMAL_PARAMETER_VIDEO_PROFILE_T encoder_param_profile = {0};
+  encoder_param_profile.hdr.id = MMAL_PARAMETER_PROFILE;
+  encoder_param_profile.hdr.size = sizeof(encoder_param_profile);
+  encoder_param_profile.profile[0].profile = MMAL_VIDEO_PROFILE_H264_BASELINE;
+  encoder_param_profile.profile[0].level = MMAL_VIDEO_LEVEL_H264_42;
+  CHK(mmal_port_parameter_set(encoder->port_out, &encoder_param_profile.hdr), "Failed to set encoder profile param");
+
+  CHK(mmal_port_parameter_set_uint32(encoder->port_out, MMAL_PARAMETER_INTRAPERIOD, params.key_frame_interval),
+      "Failed to set intra period param");
+
+  MMAL_PARAMETER_VIDEO_RATECONTROL_T encoder_param_rate_control = {0};
+  encoder_param_rate_control.hdr.id = MMAL_PARAMETER_RATECONTROL;
+  encoder_param_rate_control.hdr.size = sizeof(encoder_param_rate_control);
+  encoder_param_rate_control.control = MMAL_VIDEO_RATECONTROL_VARIABLE;
+  CHK(mmal_port_parameter_set(encoder->port_out, &encoder_param_rate_control.hdr), "Failed to set rate control param");
+
+  // Some decoders expect SPS/PPS headers to be added to every frame
+  CHK(mmal_port_parameter_set_boolean(encoder->port_out, MMAL_PARAMETER_VIDEO_ENCODE_INLINE_HEADER, MMAL_TRUE),
+      "Failed to set inline header param");
+
+  CHK(mmal_port_parameter_set_boolean(encoder->port_out, MMAL_PARAMETER_VIDEO_ENCODE_HEADERS_WITH_FRAME, MMAL_TRUE),
+      "Failed to set headers with frame param");
+
+  /* FIXME: Somehow this flag is broken? When this flag is on, the encoder will get stuck.
+  // Since our use case is mainly for real time streaming, the encoder should optimized for low latency
+  CHK(mmal_port_parameter_set_boolean(encoder->port_out, MMAL_PARAMETER_VIDEO_ENCODE_H264_LOW_LATENCY, MMAL_TRUE),
+      "Failed to set low latency param");
+  */
+
+  // Now we know the format of both ports and the requirements of the encoder, we can create
+  // our buffer headers and their associated memory buffers. We use the buffer pool API for this.
+  encoder->port_in->buffer_num = encoder->port_in->buffer_num_min;
+  // mmal calculates recommended size that's big enough to store all of the pixels
+  encoder->port_in->buffer_size = encoder->port_in->buffer_size_recommended;
+  encoder->pool_in = mmal_pool_create(encoder->port_in->buffer_num, encoder->port_in->buffer_size);
+  encoder->port_out->buffer_num = encoder->port_out->buffer_num_min;
+  encoder->port_out->buffer_size = encoder->port_out->buffer_size_recommended;
+  encoder->pool_out = mmal_pool_create(encoder->port_out->buffer_num, encoder->port_out->buffer_size);
+
+  // Create a queue to store our encoded video frames. The callback we will get when
+  // a frame has been encoded will put the frame into this queue.
+  encoder->queue_out = mmal_queue_create();
+  encoder->port_out->userdata = (void *)encoder->queue_out;
+
+  // Enable all the input port and the output port.
+  // The callback specified here is the function which will be called when the buffer header
+  // we sent to the component has been processed.
+  CHK(mmal_port_enable(encoder->port_in, encoder_in_cb), "Failed to enable input port");
+  CHK(mmal_port_enable(encoder->port_out, encoder_out_cb), "Failed to enable output port");
+
+  // Enable the component. Components will only process data when they are enabled.
+  CHK(mmal_component_enable(encoder->component), "Failed to enable component");
+
+CleanUp:
+
+  if (status.code != MMAL_SUCCESS) {
+    if (created) {
+      enc_close(encoder);
+    }
+  }
+
+  return status;
+}
+
+// enc_encode encodes y, cb, cr. The encoded frame is going to be stored in encoded_buffer.
+// IMPORTANT: the caller is responsible to release the ownership of encoded_buffer
+Status enc_encode(Encoder *encoder, Slice y, Slice cb, Slice cr, MMAL_BUFFER_HEADER_T **encoded_buffer) {
+  Status status = {0};
+  MMAL_BUFFER_HEADER_T *buffer;
+  uint32_t required_size;
+
+  // buffer should always be available since the encoding process is blocking
+  buffer = mmal_queue_get(encoder->pool_in->queue);
+  assert(buffer != NULL);
+  // buffer->data should've been allocated with enough memory to contain a frame by pool_in
+  required_size = y.len + cb.len + cr.len;
+  assert(buffer->alloc_size >= required_size);
+  memcpy(buffer->data, y.data, y.len);
+  memcpy(buffer->data + y.len, cb.data, cb.len);
+  memcpy(buffer->data + y.len + cb.len, cr.data, cr.len);
+  buffer->length = required_size;
+  CHK(mmal_port_send_buffer(encoder->port_in, buffer), "Failed to send filled buffer to input port");
+
+  while (1) {
+    // Send empty buffers to the output port to allow the encoder to start
+    // producing frames as soon as it gets input data
+    while ((buffer = mmal_queue_get(encoder->pool_out->queue)) != NULL) {
+      CHK(mmal_port_send_buffer(encoder->port_out, buffer), "Failed to send empty buffers to output port");
+    }
+
+    while ((buffer = mmal_queue_wait(encoder->queue_out)) != NULL) {
+      if ((buffer->flags & MMAL_BUFFER_HEADER_FLAG_FRAME_END) != 0) {
+        *encoded_buffer = buffer;
+        goto CleanUp;
+      }
+
+      mmal_buffer_header_release(buffer);
+    }
+  }
+
+CleanUp:
+
+  return status;
+}
+
+Status enc_close(Encoder *encoder) {
+  Status status = {0};
+
+  mmal_pool_destroy(encoder->pool_out);
+  mmal_pool_destroy(encoder->pool_in);
+  mmal_queue_destroy(encoder->queue_out);
+  mmal_component_destroy(encoder->component);
+
+CleanUp:
+
+  return status;
+}

pkg/codec/mmal/mmal.go

@@ -0,0 +1,112 @@
+// Package mmal implements a hardware accelerated H264 encoder for raspberry pi.
+// This package requires libmmal headers and libraries to be built.
+// Reference: https://github.com/raspberrypi/userland/tree/master/interface/mmal
+package mmal
+
+// #cgo pkg-config: mmal
+// #include "bridge.h"
+import "C"
+import (
+	"fmt"
+	"image"
+	"io"
+	"sync"
+	"unsafe"
+
+	"github.com/pion/mediadevices/pkg/codec"
+	"github.com/pion/mediadevices/pkg/io/video"
+	"github.com/pion/mediadevices/pkg/prop"
+)
+
+type encoder struct {
+	engine C.Encoder
+	r      video.Reader
+	mu     sync.Mutex
+	closed bool
+	cntr   int
+}
+
+func statusToErr(status *C.Status) error {
+	return fmt.Errorf("(status = %d) %s", int(status.code), C.GoString(status.msg))
+}
+
+func newEncoder(r video.Reader, p prop.Media, params Params) (codec.ReadCloser, error) {
+	if params.KeyFrameInterval == 0 {
+		params.KeyFrameInterval = 60
+	}
+
+	if params.BitRate == 0 {
+		params.BitRate = 300000
+	}
+
+	e := encoder{
+		r: video.ToI420(r),
+	}
+	status := C.enc_new(C.Params{
+		width:              C.int(p.Width),
+		height:             C.int(p.Height),
+		bitrate:            C.uint(params.BitRate),
+		key_frame_interval: C.uint(params.KeyFrameInterval),
+	}, &e.engine)
+	if status.code != 0 {
+		return nil, statusToErr(&status)
+	}
+
+	return &e, nil
+}
+
+func (e *encoder) Read() ([]byte, func(), error) {
+	e.mu.Lock()
+	defer e.mu.Unlock()
+
+	if e.closed {
+		return nil, func() {}, io.EOF
+	}
+
+	img, _, err := e.r.Read()
+	if err != nil {
+		return nil, func() {}, err
+	}
+	imgReal := img.(*image.YCbCr)
+	var y, cb, cr C.Slice
+	y.data = (*C.uchar)(&imgReal.Y[0])
+	y.len = C.int(len(imgReal.Y))
+	cb.data = (*C.uchar)(&imgReal.Cb[0])
+	cb.len = C.int(len(imgReal.Cb))
+	cr.data = (*C.uchar)(&imgReal.Cr[0])
+	cr.len = C.int(len(imgReal.Cr))
+
+	var encodedBuffer *C.MMAL_BUFFER_HEADER_T
+	status := C.enc_encode(&e.engine, y, cb, cr, &encodedBuffer)
+	if status.code != 0 {
+		return nil, func() {}, statusToErr(&status)
+	}
+
+	// GoBytes copies the C array to Go slice. After this, it's safe to release the C array
+	encoded := C.GoBytes(unsafe.Pointer(encodedBuffer.data), C.int(encodedBuffer.length))
+	// Release the buffer so that mmal can reuse this memory
+	C.mmal_buffer_header_release(encodedBuffer)
+
+	return encoded, func() {}, err
+}
+
+func (e *encoder) SetBitRate(b int) error {
+	panic("SetBitRate is not implemented")
+}
+
+func (e *encoder) ForceKeyFrame() error {
+	panic("ForceKeyFrame is not implemented")
+}
+
+func (e *encoder) Close() error {
+	e.mu.Lock()
+	defer e.mu.Unlock()
+
+	if e.closed {
+		return nil
+	}
+
+	e.closed = true
+	C.enc_close(&e.engine)
+	return nil
+}

pkg/codec/mmal/params.go

@@ -0,0 +1,31 @@
+package mmal
+
+import (
+	"github.com/pion/mediadevices/pkg/codec"
+	"github.com/pion/mediadevices/pkg/io/video"
+	"github.com/pion/mediadevices/pkg/prop"
+)
+
+// Params stores libmmal specific encoding parameters.
+type Params struct {
+	codec.BaseParams
+}
+
+// NewParams returns default mmal codec specific parameters.
+func NewParams() (Params, error) {
+	return Params{
+		BaseParams: codec.BaseParams{
+			KeyFrameInterval: 60,
+		},
+	}, nil
+}
+
+// RTPCodec represents the codec metadata
+func (p *Params) RTPCodec() *codec.RTPCodec {
+	return codec.NewRTPH264Codec(90000)
+}
+
+// BuildVideoEncoder builds mmal encoder with given params
+func (p *Params) BuildVideoEncoder(r video.Reader, property prop.Media) (codec.ReadCloser, error) {
+	return newEncoder(r, property, *p)
+}

pkg/codec/openh264/openh264.go

@@ -16,7 +16,6 @@ import (
 	"unsafe"
 
 	"github.com/pion/mediadevices/pkg/codec"
-	mio "github.com/pion/mediadevices/pkg/io"
 	"github.com/pion/mediadevices/pkg/io/video"
 	"github.com/pion/mediadevices/pkg/prop"
 )
@@ -24,7 +23,6 @@ import (
 type encoder struct {
 	engine *C.Encoder
 	r      video.Reader
-	buff   []byte
 
 	mu     sync.Mutex
 	closed bool
@@ -52,26 +50,17 @@ func newEncoder(r video.Reader, p prop.Media, params Params) (codec.ReadCloser,
 	}, nil
 }
 
-func (e *encoder) Read(p []byte) (n int, err error) {
+func (e *encoder) Read() ([]byte, func(), error) {
 	e.mu.Lock()
 	defer e.mu.Unlock()
 
 	if e.closed {
-		return 0, io.EOF
+		return nil, func() {}, io.EOF
 	}
 
-	if e.buff != nil {
-		n, err = mio.Copy(p, e.buff)
-		if err == nil {
-			e.buff = nil
-		}
-
-		return n, err
-	}
-
-	img, err := e.r.Read()
+	img, _, err := e.r.Read()
 	if err != nil {
-		return 0, err
+		return nil, func() {}, err
 	}
 
 	yuvImg := img.(*image.YCbCr)
@@ -85,16 +74,11 @@ func (e *encoder) Read(p []byte) (n int, err error) {
 		width:  C.int(bounds.Max.X - bounds.Min.X),
 	}, &rv)
 	if err := errResult(rv); err != nil {
-		return 0, fmt.Errorf("failed in encoding: %v", err)
+		return nil, func() {}, fmt.Errorf("failed in encoding: %v", err)
 	}
 
 	encoded := C.GoBytes(unsafe.Pointer(s.data), s.data_len)
-	n, err = mio.Copy(p, encoded)
-	if err != nil {
-		e.buff = encoded
-	}
-
-	return n, err
+	return encoded, func() {}, nil
 }
 
 func (e *encoder) SetBitRate(b int) error {

pkg/codec/openh264/params.go

@@ -4,7 +4,6 @@ import (
 	"github.com/pion/mediadevices/pkg/codec"
 	"github.com/pion/mediadevices/pkg/io/video"
 	"github.com/pion/mediadevices/pkg/prop"
-	"github.com/pion/webrtc/v2"
 )
 
 // Params stores libopenh264 specific encoding parameters.
@@ -21,9 +20,9 @@ func NewParams() (Params, error) {
 	}, nil
 }
 
-// Name represents the codec name
-func (p *Params) Name() string {
-	return webrtc.H264
+// RTPCodec represents the codec metadata
+func (p *Params) RTPCodec() *codec.RTPCodec {
+	return codec.NewRTPH264Codec(90000)
 }
 
 // BuildVideoEncoder builds openh264 encoder with given params

pkg/codec/opus/opus.go

@@ -72,27 +72,22 @@ func newEncoder(r audio.Reader, p prop.Media, params Params) (codec.ReadCloser,
 	return &e, nil
 }
 
-func (e *encoder) Read(p []byte) (int, error) {
-	buff, err := e.reader.Read()
+func (e *encoder) Read() ([]byte, func(), error) {
+	buff, _, err := e.reader.Read()
 	if err != nil {
-		return 0, err
+		return nil, func() {}, err
 	}
 
+	encoded := make([]byte, 1024)
 	switch b := buff.(type) {
 	case *wave.Int16Interleaved:
-		n, err := e.engine.Encode(b.Data, p)
-		if err != nil {
-			return n, err
-		}
-		return n, nil
+		n, err := e.engine.Encode(b.Data, encoded)
+		return encoded[:n:n], func() {}, err
 	case *wave.Float32Interleaved:
-		n, err := e.engine.EncodeFloat32(b.Data, p)
-		if err != nil {
-			return n, err
-		}
-		return n, nil
+		n, err := e.engine.EncodeFloat32(b.Data, encoded)
+		return encoded[:n:n], func() {}, err
 	default:
-		return 0, errors.New("unknown type of audio buffer")
+		return nil, func() {}, errors.New("unknown type of audio buffer")
 	}
 }
 

pkg/codec/opus/params.go

@@ -5,7 +5,6 @@ import (
 	"github.com/pion/mediadevices/pkg/io/audio"
 	"github.com/pion/mediadevices/pkg/prop"
 	"github.com/pion/mediadevices/pkg/wave/mixer"
-	"github.com/pion/webrtc/v2"
 )
 
 // Params stores opus specific encoding parameters.
@@ -20,9 +19,9 @@ func NewParams() (Params, error) {
 	return Params{}, nil
 }
 
-// Name represents the codec name
-func (p *Params) Name() string {
-	return webrtc.Opus
+// RTPCodec represents the codec metadata
+func (p *Params) RTPCodec() *codec.RTPCodec {
+	return codec.NewRTPOpusCodec(48000)
 }
 
 // BuildAudioEncoder builds opus encoder with given params

pkg/codec/vaapi/params.go

@@ -4,7 +4,6 @@ import (
 	"github.com/pion/mediadevices/pkg/codec"
 	"github.com/pion/mediadevices/pkg/io/video"
 	"github.com/pion/mediadevices/pkg/prop"
-	"github.com/pion/webrtc/v2"
 )
 
 // ParamsVP8 stores VP8 encoding parameters.
@@ -44,9 +43,9 @@ func NewVP8Params() (ParamsVP8, error) {
 	}, nil
 }
 
-// Name represents the codec name
-func (p *ParamsVP8) Name() string {
-	return webrtc.VP8
+// RTPCodec represents the codec metadata
+func (p *ParamsVP8) RTPCodec() *codec.RTPCodec {
+	return codec.NewRTPVP8Codec(90000)
 }
 
 // BuildVideoEncoder builds VP8 encoder with given params
@@ -113,9 +112,9 @@ func NewVP9Params() (ParamsVP9, error) {
 	}, nil
 }
 
-// Name represents the codec name
-func (p *ParamsVP9) Name() string {
-	return webrtc.VP9
+// RTPCodec represents the codec metadata
+func (p *ParamsVP9) RTPCodec() *codec.RTPCodec {
+	return codec.NewRTPVP9Codec(90000)
 }
 
 // BuildVideoEncoder builds VP9 encoder with given params

pkg/codec/vaapi/vp8.go

@@ -64,7 +64,6 @@ import (
 	"unsafe"
 
 	"github.com/pion/mediadevices/pkg/codec"
-	mio "github.com/pion/mediadevices/pkg/io"
 	"github.com/pion/mediadevices/pkg/io/video"
 	"github.com/pion/mediadevices/pkg/prop"
 )
@@ -80,7 +79,6 @@ const (
 
 type encoderVP8 struct {
 	r     video.Reader
-	buf   []byte
 	frame []byte
 
 	fdDRI    C.int
@@ -297,25 +295,17 @@ func newVP8Encoder(r video.Reader, p prop.Media, params ParamsVP8) (codec.ReadCl
 	return e, nil
 }
 
-func (e *encoderVP8) Read(p []byte) (int, error) {
+func (e *encoderVP8) Read() ([]byte, func(), error) {
 	e.mu.Lock()
 	defer e.mu.Unlock()
 
 	if e.closed {
-		return 0, io.EOF
+		return nil, func() {}, io.EOF
 	}
 
-	if e.buf != nil {
-		n, err := mio.Copy(p, e.buf)
-		if err == nil {
-			e.buf = nil
-		}
-		return n, err
-	}
-
-	img, err := e.r.Read()
+	img, _, err := e.r.Read()
 	if err != nil {
-		return 0, err
+		return nil, func() {}, err
 	}
 	yuvImg := img.(*image.YCbCr)
 
@@ -357,7 +347,7 @@ func (e *encoderVP8) Read(p []byte) (int, error) {
 			}
 		}
 		if e.picParam.reconstructed_frame == C.VA_INVALID_SURFACE {
-			return 0, errors.New("no available surface")
+			return nil, func() {}, errors.New("no available surface")
 		}
 
 		C.setForceKFFlagVP8(&e.picParam, 0)
@@ -425,7 +415,7 @@ func (e *encoderVP8) Read(p []byte) (int, error) {
 			C.size_t(uintptr(p.src)),
 			&id,
 		); s != C.VA_STATUS_SUCCESS {
-			return 0, fmt.Errorf("failed to create buffer: %s", C.GoString(C.vaErrorStr(s)))
+			return nil, func() {}, fmt.Errorf("failed to create buffer: %s", C.GoString(C.vaErrorStr(s)))
 		}
 		buffs = append(buffs, id)
 	}
@@ -435,17 +425,17 @@ func (e *encoderVP8) Read(p []byte) (int, error) {
 		e.display, e.ctxID,
 		e.surfs[surfaceVP8Input],
 	); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to begin picture: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to begin picture: %s", C.GoString(C.vaErrorStr(s)))
 	}
 
 	// Upload image
 	var vaImg C.VAImage
 	var rawBuf unsafe.Pointer
 	if s := C.vaDeriveImage(e.display, e.surfs[surfaceVP8Input], &vaImg); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to derive image: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to derive image: %s", C.GoString(C.vaErrorStr(s)))
 	}
 	if s := C.vaMapBuffer(e.display, vaImg.buf, &rawBuf); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to map buffer: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to map buffer: %s", C.GoString(C.vaErrorStr(s)))
 	}
 	// TODO: use vaImg.pitches to support padding
 	C.memcpy(
@@ -461,10 +451,10 @@ func (e *encoderVP8) Read(p []byte) (int, error) {
 		unsafe.Pointer(&yuvImg.Cr[0]), C.size_t(len(yuvImg.Cr)),
 	)
 	if s := C.vaUnmapBuffer(e.display, vaImg.buf); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to unmap buffer: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to unmap buffer: %s", C.GoString(C.vaErrorStr(s)))
 	}
 	if s := C.vaDestroyImage(e.display, vaImg.image_id); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to destroy image: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to destroy image: %s", C.GoString(C.vaErrorStr(s)))
 	}
 
 	if s := C.vaRenderPicture(
@@ -472,38 +462,38 @@ func (e *encoderVP8) Read(p []byte) (int, error) {
 		&buffs[1], // 0 is for ouput
 		C.int(len(buffs)-1),
 	); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to render picture: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to render picture: %s", C.GoString(C.vaErrorStr(s)))
 	}
 	if s := C.vaEndPicture(
 		e.display, e.ctxID,
 	); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to end picture: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to end picture: %s", C.GoString(C.vaErrorStr(s)))
 	}
 
 	// Load encoded data
 	for retry := 3; retry >= 0; retry-- {
 		if s := C.vaSyncSurface(e.display, e.picParam.reconstructed_frame); s != C.VA_STATUS_SUCCESS {
-			return 0, fmt.Errorf("failed to sync surface: %s", C.GoString(C.vaErrorStr(s)))
+			return nil, func() {}, fmt.Errorf("failed to sync surface: %s", C.GoString(C.vaErrorStr(s)))
 		}
 		var surfStat C.VASurfaceStatus
 		if s := C.vaQuerySurfaceStatus(
 			e.display, e.picParam.reconstructed_frame, &surfStat,
 		); s != C.VA_STATUS_SUCCESS {
-			return 0, fmt.Errorf("failed to query surface status: %s", C.GoString(C.vaErrorStr(s)))
+			return nil, func() {}, fmt.Errorf("failed to query surface status: %s", C.GoString(C.vaErrorStr(s)))
 		}
 		if surfStat == C.VASurfaceReady {
 			break
 		}
 		if retry == 0 {
-			return 0, fmt.Errorf("failed to sync surface: %d", surfStat)
+			return nil, func() {}, fmt.Errorf("failed to sync surface: %d", surfStat)
 		}
 	}
 	var seg *C.VACodedBufferSegment
 	if s := C.vaMapBufferSeg(e.display, buffs[0], &seg); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to map buffer: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to map buffer: %s", C.GoString(C.vaErrorStr(s)))
 	}
 	if seg.status&C.VA_CODED_BUF_STATUS_SLICE_OVERFLOW_MASK != 0 {
-		return 0, errors.New("buffer size too small")
+		return nil, func() {}, errors.New("buffer size too small")
 	}
 
 	if cap(e.frame) < int(seg.size) {
@@ -516,13 +506,13 @@ func (e *encoderVP8) Read(p []byte) (int, error) {
 	)
 
 	if s := C.vaUnmapBuffer(e.display, buffs[0]); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to unmap buffer: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to unmap buffer: %s", C.GoString(C.vaErrorStr(s)))
 	}
 
 	// Destroy buffers
 	for _, b := range buffs {
 		if s := C.vaDestroyBuffer(e.display, b); s != C.VA_STATUS_SUCCESS {
-			return 0, fmt.Errorf("failed to destroy buffer: %s", C.GoString(C.vaErrorStr(s)))
+			return nil, func() {}, fmt.Errorf("failed to destroy buffer: %s", C.GoString(C.vaErrorStr(s)))
 		}
 	}
 
@@ -545,11 +535,9 @@ func (e *encoderVP8) Read(p []byte) (int, error) {
 	e.picParam.ref_last_frame = e.picParam.reconstructed_frame
 	C.setRefreshLastFlagVP8(&e.picParam, 1)
 
-	n, err := mio.Copy(p, e.frame)
-	if err != nil {
-		e.buf = e.frame
-	}
-	return n, err
+	encoded := make([]byte, len(e.frame))
+	copy(encoded, e.frame)
+	return encoded, func() {}, err
 }
 
 func (e *encoderVP8) SetBitRate(b int) error {

pkg/codec/vaapi/vp9.go

@@ -47,7 +47,6 @@ import (
 	"unsafe"
 
 	"github.com/pion/mediadevices/pkg/codec"
-	mio "github.com/pion/mediadevices/pkg/io"
 	"github.com/pion/mediadevices/pkg/io/video"
 	"github.com/pion/mediadevices/pkg/prop"
 )
@@ -67,7 +66,6 @@ const (
 
 type encoderVP9 struct {
 	r     video.Reader
-	buf   []byte
 	frame []byte
 
 	fdDRI    C.int
@@ -286,25 +284,17 @@ func newVP9Encoder(r video.Reader, p prop.Media, params ParamsVP9) (codec.ReadCl
 	return e, nil
 }
 
-func (e *encoderVP9) Read(p []byte) (int, error) {
+func (e *encoderVP9) Read() ([]byte, func(), error) {
 	e.mu.Lock()
 	defer e.mu.Unlock()
 
 	if e.closed {
-		return 0, io.EOF
+		return nil, func() {}, io.EOF
 	}
 
-	if e.buf != nil {
-		n, err := mio.Copy(p, e.buf)
-		if err == nil {
-			e.buf = nil
-		}
-		return n, err
-	}
-
-	img, err := e.r.Read()
+	img, _, err := e.r.Read()
 	if err != nil {
-		return 0, err
+		return nil, func() {}, err
 	}
 	yuvImg := img.(*image.YCbCr)
 
@@ -388,7 +378,7 @@ func (e *encoderVP9) Read(p []byte) (int, error) {
 			C.size_t(uintptr(p.src)),
 			&id,
 		); s != C.VA_STATUS_SUCCESS {
-			return 0, fmt.Errorf("failed to create buffer: %s", C.GoString(C.vaErrorStr(s)))
+			return nil, func() {}, fmt.Errorf("failed to create buffer: %s", C.GoString(C.vaErrorStr(s)))
 		}
 		buffs = append(buffs, id)
 	}
@@ -398,17 +388,17 @@ func (e *encoderVP9) Read(p []byte) (int, error) {
 		e.display, e.ctxID,
 		e.surfs[surfaceVP9Input],
 	); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to begin picture: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to begin picture: %s", C.GoString(C.vaErrorStr(s)))
 	}
 
 	// Upload image
 	var vaImg C.VAImage
 	var rawBuf unsafe.Pointer
 	if s := C.vaDeriveImage(e.display, e.surfs[surfaceVP9Input], &vaImg); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to derive image: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to derive image: %s", C.GoString(C.vaErrorStr(s)))
 	}
 	if s := C.vaMapBuffer(e.display, vaImg.buf, &rawBuf); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to map buffer: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to map buffer: %s", C.GoString(C.vaErrorStr(s)))
 	}
 	// TODO: use vaImg.pitches to support padding
 	C.copyI420toNV12(
@@ -419,10 +409,10 @@ func (e *encoderVP9) Read(p []byte) (int, error) {
 		C.uint(len(yuvImg.Y)),
 	)
 	if s := C.vaUnmapBuffer(e.display, vaImg.buf); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to unmap buffer: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to unmap buffer: %s", C.GoString(C.vaErrorStr(s)))
 	}
 	if s := C.vaDestroyImage(e.display, vaImg.image_id); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to destroy image: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to destroy image: %s", C.GoString(C.vaErrorStr(s)))
 	}
 
 	if s := C.vaRenderPicture(
@@ -430,27 +420,27 @@ func (e *encoderVP9) Read(p []byte) (int, error) {
 		&buffs[1], // 0 is for ouput
 		C.int(len(buffs)-1),
 	); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to render picture: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to render picture: %s", C.GoString(C.vaErrorStr(s)))
 	}
 	if s := C.vaEndPicture(
 		e.display, e.ctxID,
 	); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to end picture: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to end picture: %s", C.GoString(C.vaErrorStr(s)))
 	}
 
 	// Load encoded data
 	if s := C.vaSyncSurface(e.display, e.picParam.reconstructed_frame); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to sync surface: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to sync surface: %s", C.GoString(C.vaErrorStr(s)))
 	}
 	var surfStat C.VASurfaceStatus
 	if s := C.vaQuerySurfaceStatus(
 		e.display, e.picParam.reconstructed_frame, &surfStat,
 	); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to query surface status: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to query surface status: %s", C.GoString(C.vaErrorStr(s)))
 	}
 	var seg *C.VACodedBufferSegment
 	if s := C.vaMapBufferSeg(e.display, buffs[0], &seg); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to map buffer: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to map buffer: %s", C.GoString(C.vaErrorStr(s)))
 	}
 	if cap(e.frame) < int(seg.size) {
 		e.frame = make([]byte, int(seg.size))
@@ -462,13 +452,13 @@ func (e *encoderVP9) Read(p []byte) (int, error) {
 	)
 
 	if s := C.vaUnmapBuffer(e.display, buffs[0]); s != C.VA_STATUS_SUCCESS {
-		return 0, fmt.Errorf("failed to unmap buffer: %s", C.GoString(C.vaErrorStr(s)))
+		return nil, func() {}, fmt.Errorf("failed to unmap buffer: %s", C.GoString(C.vaErrorStr(s)))
 	}
 
 	// Destroy buffers
 	for _, b := range buffs {
 		if s := C.vaDestroyBuffer(e.display, b); s != C.VA_STATUS_SUCCESS {
-			return 0, fmt.Errorf("failed to destroy buffer: %s", C.GoString(C.vaErrorStr(s)))
+			return nil, func() {}, fmt.Errorf("failed to destroy buffer: %s", C.GoString(C.vaErrorStr(s)))
 		}
 	}
 
@@ -480,11 +470,9 @@ func (e *encoderVP9) Read(p []byte) (int, error) {
 		e.slotCurr = 0
 	}
 
-	n, err := mio.Copy(p, e.frame)
-	if err != nil {
-		e.buf = e.frame
-	}
-	return n, err
+	encoded := make([]byte, len(e.frame))
+	copy(encoded, e.frame)
+	return encoded, func() {}, err
 }
 
 func (e *encoderVP9) SetBitRate(b int) error {

pkg/codec/vpx/vpx.go

@@ -56,10 +56,8 @@ import (
 	"unsafe"
 
 	"github.com/pion/mediadevices/pkg/codec"
-	mio "github.com/pion/mediadevices/pkg/io"
 	"github.com/pion/mediadevices/pkg/io/video"
 	"github.com/pion/mediadevices/pkg/prop"
-	"github.com/pion/webrtc/v2"
 )
 
 type encoder struct {
@@ -68,7 +66,6 @@ type encoder struct {
 	cfg        *C.vpx_codec_enc_cfg_t
 	r          video.Reader
 	frameIndex int
-	buff       []byte
 	tStart     int
 	tLastFrame int
 	frame      []byte
@@ -95,9 +92,9 @@ func NewVP8Params() (VP8Params, error) {
 	}, nil
 }
 
-// Name represents the codec name
-func (p *VP8Params) Name() string {
-	return webrtc.VP8
+// RTPCodec represents the codec metadata
+func (p *VP8Params) RTPCodec() *codec.RTPCodec {
+	return codec.NewRTPVP8Codec(90000)
 }
 
 // BuildVideoEncoder builds VP8 encoder with given params
@@ -122,9 +119,9 @@ func NewVP9Params() (VP9Params, error) {
 	}, nil
 }
 
-// Name represents the codec name
-func (p *VP9Params) Name() string {
-	return webrtc.VP9
+// RTPCodec represents the codec metadata
+func (p *VP9Params) RTPCodec() *codec.RTPCodec {
+	return codec.NewRTPVP9Codec(90000)
 }
 
 // BuildVideoEncoder builds VP9 encoder with given params
@@ -207,25 +204,17 @@ func newEncoder(r video.Reader, p prop.Media, params Params, codecIface *C.vpx_c
 	}, nil
 }
 
-func (e *encoder) Read(p []byte) (int, error) {
+func (e *encoder) Read() ([]byte, func(), error) {
 	e.mu.Lock()
 	defer e.mu.Unlock()
 
 	if e.closed {
-		return 0, io.EOF
+		return nil, func() {}, io.EOF
 	}
 
-	if e.buff != nil {
-		n, err := mio.Copy(p, e.buff)
-		if err == nil {
-			e.buff = nil
-		}
-		return n, err
-	}
-
-	img, err := e.r.Read()
+	img, _, err := e.r.Read()
 	if err != nil {
-		return 0, err
+		return nil, func() {}, err
 	}
 	yuvImg := img.(*image.YCbCr)
 	bounds := yuvImg.Bounds()
@@ -241,7 +230,7 @@ func (e *encoder) Read(p []byte) (int, error) {
 	if e.cfg.g_w != C.uint(width) || e.cfg.g_h != C.uint(height) {
 		e.cfg.g_w, e.cfg.g_h = C.uint(width), C.uint(height)
 		if ec := C.vpx_codec_enc_config_set(e.codec, e.cfg); ec != C.VPX_CODEC_OK {
-			return 0, fmt.Errorf("vpx_codec_enc_config_set failed (%d)", ec)
+			return nil, func() {}, fmt.Errorf("vpx_codec_enc_config_set failed (%d)", ec)
 		}
 		e.raw.w, e.raw.h = C.uint(width), C.uint(height)
 		e.raw.r_w, e.raw.r_h = C.uint(width), C.uint(height)
@@ -254,7 +243,7 @@ func (e *encoder) Read(p []byte) (int, error) {
 		C.long(t-e.tStart), C.ulong(t-e.tLastFrame), C.long(flags), C.ulong(e.deadline),
 		(*C.uchar)(&yuvImg.Y[0]), (*C.uchar)(&yuvImg.Cb[0]), (*C.uchar)(&yuvImg.Cr[0]),
 	); ec != C.VPX_CODEC_OK {
-		return 0, fmt.Errorf("vpx_codec_encode failed (%d)", ec)
+		return nil, func() {}, fmt.Errorf("vpx_codec_encode failed (%d)", ec)
 	}
 
 	e.frameIndex++
@@ -272,11 +261,10 @@ func (e *encoder) Read(p []byte) (int, error) {
 			e.frame = append(e.frame, encoded...)
 		}
 	}
-	n, err := mio.Copy(p, e.frame)
-	if err != nil {
-		e.buff = e.frame
-	}
-	return n, err
+
+	encoded := make([]byte, len(e.frame))
+	copy(encoded, e.frame)
+	return encoded, func() {}, err
 }
 
 func (e *encoder) SetBitRate(b int) error {

pkg/codec/x264/bridge.h

@@ -47,7 +47,7 @@ Encoder *enc_new(x264_param_t param, char *preset, int *rc) {
   e->param.b_repeat_headers = 1;
   e->param.b_annexb = 1;
 
-  if (x264_param_apply_profile(&e->param, "baseline") < 0) {
+  if (x264_param_apply_profile(&e->param, "high") < 0) {
     *rc = ERR_APPLY_PROFILE;
     goto fail;
   }
@@ -95,4 +95,4 @@ void enc_close(Encoder *e, int *rc) {
   x264_encoder_close(e->h);
   x264_picture_clean(&e->pic_in);
   free(e);
-}
+}

pkg/codec/x264/params.go

@@ -4,7 +4,6 @@ import (
 	"github.com/pion/mediadevices/pkg/codec"
 	"github.com/pion/mediadevices/pkg/io/video"
 	"github.com/pion/mediadevices/pkg/prop"
-	"github.com/pion/webrtc/v2"
 )
 
 // Params stores libx264 specific encoding parameters.
@@ -40,9 +39,9 @@ func NewParams() (Params, error) {
 	}, nil
 }
 
-// Name represents the codec name
-func (p *Params) Name() string {
-	return webrtc.H264
+// RTPCodec represents the codec metadata
+func (p *Params) RTPCodec() *codec.RTPCodec {
+	return codec.NewRTPH264Codec(90000)
 }
 
 // BuildVideoEncoder builds x264 encoder with given params

pkg/codec/x264/x264.go

@@ -14,14 +14,12 @@ import (
 	"unsafe"
 
 	"github.com/pion/mediadevices/pkg/codec"
-	mio "github.com/pion/mediadevices/pkg/io"
 	"github.com/pion/mediadevices/pkg/io/video"
 	"github.com/pion/mediadevices/pkg/prop"
 )
 
 type encoder struct {
 	engine *C.Encoder
-	buff   []byte
 	r      video.Reader
 	mu     sync.Mutex
 	closed bool
@@ -96,25 +94,17 @@ func newEncoder(r video.Reader, p prop.Media, params Params) (codec.ReadCloser,
 	return &e, nil
 }
 
-func (e *encoder) Read(p []byte) (int, error) {
+func (e *encoder) Read() ([]byte, func(), error) {
 	e.mu.Lock()
 	defer e.mu.Unlock()
 
 	if e.closed {
-		return 0, io.EOF
+		return nil, func() {}, io.EOF
 	}
 
-	if e.buff != nil {
-		n, err := mio.Copy(p, e.buff)
-		if err == nil {
-			e.buff = nil
-		}
-		return n, err
-	}
-
-	img, err := e.r.Read()
+	img, _, err := e.r.Read()
 	if err != nil {
-		return 0, err
+		return nil, func() {}, err
 	}
 	yuvImg := img.(*image.YCbCr)
 
@@ -127,15 +117,11 @@ func (e *encoder) Read(p []byte) (int, error) {
 		&rc,
 	)
 	if err := errFromC(rc); err != nil {
-		return 0, err
+		return nil, func() {}, err
 	}
 
 	encoded := C.GoBytes(unsafe.Pointer(s.data), s.data_len)
-	n, err := mio.Copy(p, encoded)
-	if err != nil {
-		e.buff = encoded
-	}
-	return n, err
+	return encoded, func() {}, err
 }
 
 func (e *encoder) SetBitRate(b int) error {

pkg/driver/audiotest/dummy.go

@@ -52,10 +52,10 @@ func (d *dummy) AudioRecord(p prop.Media) (audio.Reader, error) {
 
 	closed := d.closed
 
-	reader := audio.ReaderFunc(func() (wave.Audio, error) {
+	reader := audio.ReaderFunc(func() (wave.Audio, func(), error) {
 		select {
 		case <-closed:
-			return nil, io.EOF
+			return nil, func() {}, io.EOF
 		default:
 		}
 
@@ -78,7 +78,7 @@ func (d *dummy) AudioRecord(p prop.Media) (audio.Reader, error) {
 				a.SetFloat32(i, ch, wave.Float32Sample(sin[phase]))
 			}
 		}
-		return a, nil
+		return a, func() {}, nil
 	})
 	return reader, nil
 }

pkg/driver/camera/camera_darwin.go

@@ -56,10 +56,10 @@ func (cam *camera) VideoRecord(property prop.Media) (video.Reader, error) {
 	if err != nil {
 		return nil, err
 	}
-	r := video.ReaderFunc(func() (image.Image, error) {
-		frame, err := rc.Read()
+	r := video.ReaderFunc(func() (image.Image, func(), error) {
+		frame, _, err := rc.Read()
 		if err != nil {
-			return nil, err
+			return nil, func() {}, err
 		}
 		return decoder.Decode(frame, property.Width, property.Height)
 	})

pkg/driver/camera/camera_linux.go

@@ -8,7 +8,8 @@ import (
 	"errors"
 	"image"
 	"io"
-	"io/ioutil"
+	"os"
+	"path/filepath"
 	"sync"
 
 	"github.com/blackjack/webcam"
@@ -25,6 +26,36 @@ const (
 var (
 	errReadTimeout = errors.New("read timeout")
 	errEmptyFrame  = errors.New("empty frame")
+	// Reference: https://commons.wikimedia.org/wiki/File:Vector_Video_Standards2.svg
+	supportedResolutions = [][2]int{
+		{320, 240},
+		{640, 480},
+		{768, 576},
+		{800, 600},
+		{1024, 768},
+		{1280, 854},
+		{1280, 960},
+		{1280, 1024},
+		{1400, 1050},
+		{1600, 1200},
+		{2048, 1536},
+		{320, 200},
+		{800, 480},
+		{854, 480},
+		{1024, 600},
+		{1152, 768},
+		{1280, 720},
+		{1280, 768},
+		{1366, 768},
+		{1280, 800},
+		{1440, 900},
+		{1440, 960},
+		{1680, 1050},
+		{1920, 1080},
+		{2048, 1080},
+		{1920, 1200},
+		{2560, 1600},
+	}
 )
 
 // Camera implementation using v4l2
@@ -40,19 +71,38 @@ type camera struct {
 }
 
 func init() {
-	searchPath := "/dev/v4l/by-path/"
-	devices, err := ioutil.ReadDir(searchPath)
-	if err != nil {
-		// No v4l device.
-		return
-	}
-	for _, device := range devices {
-		cam := newCamera(searchPath + device.Name())
-		driver.GetManager().Register(cam, driver.Info{
-			Label:      device.Name(),
-			DeviceType: driver.Camera,
-		})
+	discovered := make(map[string]struct{})
+
+	discover := func(pattern string) {
+		devices, err := filepath.Glob(pattern)
+		if err != nil {
+			// No v4l device.
+			return
+		}
+		for _, device := range devices {
+			label := filepath.Base(device)
+			reallink, err := os.Readlink(device)
+			if err != nil {
+				reallink = label
+			} else {
+				reallink = filepath.Base(reallink)
+			}
+
+			if _, ok := discovered[reallink]; ok {
+				continue
+			}
+
+			discovered[reallink] = struct{}{}
+			cam := newCamera(device)
+			driver.GetManager().Register(cam, driver.Info{
+				Label:      label,
+				DeviceType: driver.Camera,
+			})
+		}
 	}
+
+	discover("/dev/v4l/by-path/*")
+	discover("/dev/video*")
 }
 
 func newCamera(path string) *camera {
@@ -83,6 +133,8 @@ func (c *camera) Open() error {
 		return err
 	}
 
+	// Late frames should be discarded. Buffering should be handled in higher level.
+	cam.SetBufferCount(1)
 	c.cam = cam
 	return nil
 }
@@ -130,7 +182,7 @@ func (c *camera) VideoRecord(p prop.Media) (video.Reader, error) {
 	ctx, cancel := context.WithCancel(context.Background())
 	c.cancel = cancel
 	var buf []byte
-	r := video.ReaderFunc(func() (img image.Image, err error) {
+	r := video.ReaderFunc(func() (img image.Image, release func(), err error) {
 		// Lock to avoid accessing the buffer after StopStreaming()
 		c.mutex.Lock()
 		defer c.mutex.Unlock()
@@ -139,23 +191,23 @@ func (c *camera) VideoRecord(p prop.Media) (video.Reader, error) {
 		for i := 0; i < maxEmptyFrameCount; i++ {
 			if ctx.Err() != nil {
 				// Return EOF if the camera is already closed.
-				return nil, io.EOF
+				return nil, func() {}, io.EOF
 			}
 
 			err := cam.WaitForFrame(5) // 5 seconds
 			switch err.(type) {
 			case nil:
 			case *webcam.Timeout:
-				return nil, errReadTimeout
+				return nil, func() {}, errReadTimeout
 			default:
 				// Camera has been stopped.
-				return nil, err
+				return nil, func() {}, err
 			}
 
 			b, err := cam.ReadFrame()
 			if err != nil {
 				// Camera has been stopped.
-				return nil, err
+				return nil, func() {}, err
 			}
 
 			// Frame is empty.
@@ -175,7 +227,7 @@ func (c *camera) VideoRecord(p prop.Media) (video.Reader, error) {
 			n := copy(buf, b)
 			return decoder.Decode(buf[:n], p.Width, p.Height)
 		}
-		return nil, errEmptyFrame
+		return nil, func() {}, errEmptyFrame
 	})
 
 	return r, nil
@@ -185,13 +237,46 @@ func (c *camera) Properties() []prop.Media {
 	properties := make([]prop.Media, 0)
 	for format := range c.cam.GetSupportedFormats() {
 		for _, frameSize := range c.cam.GetSupportedFrameSizes(format) {
-			properties = append(properties, prop.Media{
-				Video: prop.Video{
-					Width:       int(frameSize.MaxWidth),
-					Height:      int(frameSize.MaxHeight),
-					FrameFormat: c.formats[format],
-				},
-			})
+			supportedFormat, ok := c.formats[format]
+			if !ok {
+				continue
+			}
+
+			if frameSize.StepWidth == 0 || frameSize.StepHeight == 0 {
+				properties = append(properties, prop.Media{
+					Video: prop.Video{
+						Width:       int(frameSize.MaxWidth),
+						Height:      int(frameSize.MaxHeight),
+						FrameFormat: supportedFormat,
+					},
+				})
+			} else {
+				// FIXME: we should probably use a custom data structure to capture all of the supported resolutions
+				for _, supportedResolution := range supportedResolutions {
+					minWidth, minHeight := int(frameSize.MinWidth), int(frameSize.MinHeight)
+					maxWidth, maxHeight := int(frameSize.MaxWidth), int(frameSize.MaxHeight)
+					stepWidth, stepHeight := int(frameSize.StepWidth), int(frameSize.StepHeight)
+					width, height := supportedResolution[0], supportedResolution[1]
+
+					if width < minWidth || width > maxWidth ||
+						height < minHeight || height > maxHeight {
+						continue
+					}
+
+					if (width-minWidth)%stepWidth != 0 ||
+						(height-minHeight)%stepHeight != 0 {
+						continue
+					}
+
+					properties = append(properties, prop.Media{
+						Video: prop.Video{
+							Width:       width,
+							Height:      height,
+							FrameFormat: supportedFormat,
+						},
+					})
+				}
+			}
 		}
 	}
 	return properties

pkg/driver/camera/camera_windows.go

@@ -116,10 +116,10 @@ func (c *camera) VideoRecord(p prop.Media) (video.Reader, error) {
 
 	img := &image.YCbCr{}
 
-	r := video.ReaderFunc(func() (image.Image, error) {
+	r := video.ReaderFunc(func() (image.Image, func(), error) {
 		b, ok := <-c.ch
 		if !ok {
-			return nil, io.EOF
+			return nil, func() {}, io.EOF
 		}
 		img.Y = b[:nPix]
 		img.Cb = b[nPix : nPix+nPix/2]
@@ -128,7 +128,7 @@ func (c *camera) VideoRecord(p prop.Media) (video.Reader, error) {
 		img.CStride = p.Width / 2
 		img.SubsampleRatio = image.YCbCrSubsampleRatio422
 		img.Rect = image.Rect(0, 0, p.Width, p.Height)
-		return img, nil
+		return img, func() {}, nil
 	})
 	return r, nil
 }

pkg/driver/microphone/microphone.go

@@ -1 +1,204 @@
 package microphone
+
+import (
+	"encoding/binary"
+	"errors"
+	"fmt"
+	"io"
+	"time"
+	"unsafe"
+
+	"github.com/gen2brain/malgo"
+	"github.com/pion/mediadevices/internal/logging"
+	"github.com/pion/mediadevices/pkg/driver"
+	"github.com/pion/mediadevices/pkg/io/audio"
+	"github.com/pion/mediadevices/pkg/prop"
+	"github.com/pion/mediadevices/pkg/wave"
+)
+
+const (
+	maxDeviceIDLength = 20
+	// TODO: should replace this with a more flexible approach
+	sampleRateStep    = 1000
+	initialBufferSize = 1024
+)
+
+var logger = logging.NewLogger("mediadevices/driver/microphone")
+var ctx *malgo.AllocatedContext
+var hostEndian binary.ByteOrder
+var (
+	errUnsupportedFormat = errors.New("the provided audio format is not supported")
+)
+
+type microphone struct {
+	malgo.DeviceInfo
+	chunkChan chan []byte
+}
+
+func init() {
+	var err error
+	/*
+		backends := []malgo.Backend{
+			malgo.BackendPulseaudio,
+			malgo.BackendAlsa,
+		}
+	*/
+	ctx, err = malgo.InitContext(nil, malgo.ContextConfig{}, func(message string) {
+		logger.Debugf("%v\n", message)
+	})
+	if err != nil {
+		panic(err)
+	}
+
+	devices, err := ctx.Devices(malgo.Capture)
+	if err != nil {
+		panic(err)
+	}
+
+	for _, device := range devices {
+		// TODO: Detect default device and prioritize it
+		driver.GetManager().Register(newMicrophone(device), driver.Info{
+			Label:      device.ID.String(),
+			DeviceType: driver.Microphone,
+		})
+	}
+
+	// Decide which endian
+	switch v := *(*uint16)(unsafe.Pointer(&([]byte{0x12, 0x34}[0]))); v {
+	case 0x1234:
+		hostEndian = binary.BigEndian
+	case 0x3412:
+		hostEndian = binary.LittleEndian
+	default:
+		panic(fmt.Sprintf("failed to determine host endianness: %x", v))
+	}
+}
+
+func newMicrophone(info malgo.DeviceInfo) *microphone {
+	return &microphone{
+		DeviceInfo: info,
+	}
+}
+
+func (m *microphone) Open() error {
+	m.chunkChan = make(chan []byte, 1)
+	return nil
+}
+
+func (m *microphone) Close() error {
+	if m.chunkChan != nil {
+		close(m.chunkChan)
+		m.chunkChan = nil
+	}
+	return nil
+}
+
+func (m *microphone) AudioRecord(inputProp prop.Media) (audio.Reader, error) {
+	var config malgo.DeviceConfig
+	var callbacks malgo.DeviceCallbacks
+
+	decoder, err := wave.NewDecoder(&wave.RawFormat{
+		SampleSize:  inputProp.SampleSize,
+		IsFloat:     inputProp.IsFloat,
+		Interleaved: inputProp.IsInterleaved,
+	})
+	if err != nil {
+		return nil, err
+	}
+
+	config.DeviceType = malgo.Capture
+	config.PerformanceProfile = malgo.LowLatency
+	config.Capture.Channels = uint32(inputProp.ChannelCount)
+	config.SampleRate = uint32(inputProp.SampleRate)
+	if inputProp.SampleSize == 4 && inputProp.IsFloat {
+		config.Capture.Format = malgo.FormatF32
+	} else if inputProp.SampleSize == 2 && !inputProp.IsFloat {
+		config.Capture.Format = malgo.FormatS16
+	} else {
+		return nil, errUnsupportedFormat
+	}
+
+	onRecvChunk := func(_, chunk []byte, framecount uint32) {
+		m.chunkChan <- chunk
+	}
+	callbacks.Data = onRecvChunk
+
+	device, err := malgo.InitDevice(ctx.Context, config, callbacks)
+	if err != nil {
+		return nil, err
+	}
+
+	err = device.Start()
+	if err != nil {
+		return nil, err
+	}
+
+	return audio.ReaderFunc(func() (wave.Audio, func(), error) {
+		chunk, ok := <-m.chunkChan
+		if !ok {
+			device.Stop()
+			device.Uninit()
+			return nil, func() {}, io.EOF
+		}
+
+		decodedChunk, err := decoder.Decode(hostEndian, chunk, inputProp.ChannelCount)
+		// FIXME: the decoder should also fill this information
+		decodedChunk.(*wave.Float32Interleaved).Size.SamplingRate = inputProp.SampleRate
+		return decodedChunk, func() {}, err
+	}), nil
+}
+
+func (m *microphone) Properties() []prop.Media {
+	var supportedProps []prop.Media
+	logger.Debug("Querying properties")
+
+	var isBigEndian bool
+	// miniaudio only uses the host endian
+	if hostEndian == binary.BigEndian {
+		isBigEndian = true
+	}
+
+	for ch := m.MinChannels; ch <= m.MaxChannels; ch++ {
+		for sampleRate := m.MinSampleRate; sampleRate <= m.MaxSampleRate; sampleRate += sampleRateStep {
+			for i := 0; i < int(m.FormatCount); i++ {
+				format := m.Formats[i]
+
+				supportedProp := prop.Media{
+					Audio: prop.Audio{
+						ChannelCount: int(ch),
+						SampleRate:   int(sampleRate),
+						IsBigEndian:  isBigEndian,
+						// miniaudio only supports interleaved at the moment
+						IsInterleaved: true,
+					},
+				}
+
+				switch malgo.FormatType(format) {
+				case malgo.FormatF32:
+					supportedProp.SampleSize = 4
+					supportedProp.IsFloat = true
+				case malgo.FormatS16:
+					supportedProp.SampleSize = 2
+					supportedProp.IsFloat = false
+				}
+
+				supportedProps = append(supportedProps, supportedProp)
+			}
+		}
+	}
+
+	// FIXME: remove this hardcoded value. Malgo doesn't support "ma_context_get_device_info" API yet. The above iterations
+	//        will always return nothing as of now
+	supportedProps = append(supportedProps, prop.Media{
+		Audio: prop.Audio{
+			Latency:       time.Millisecond * 20,
+			ChannelCount:  1,
+			SampleRate:    48000,
+			SampleSize:    4,
+			IsFloat:       true,
+			IsBigEndian:   isBigEndian,
+			IsInterleaved: true,
+		},
+	})
+	return supportedProps
+}

pkg/driver/microphone/microphone_linux.go

@@ -1,137 +0,0 @@
-package microphone
-
-import (
-	"io"
-	"time"
-
-	"github.com/jfreymuth/pulse"
-	"github.com/pion/mediadevices/pkg/driver"
-	"github.com/pion/mediadevices/pkg/io/audio"
-	"github.com/pion/mediadevices/pkg/prop"
-	"github.com/pion/mediadevices/pkg/wave"
-)
-
-type microphone struct {
-	c           *pulse.Client
-	id          string
-	samplesChan chan<- []int16
-}
-
-func init() {
-	pa, err := pulse.NewClient()
-	if err != nil {
-		// No pulseaudio
-		return
-	}
-	defer pa.Close()
-	sources, err := pa.ListSources()
-	if err != nil {
-		panic(err)
-	}
-	defaultSource, err := pa.DefaultSource()
-	if err != nil {
-		panic(err)
-	}
-	for _, source := range sources {
-		priority := driver.PriorityNormal
-		if defaultSource.ID() == source.ID() {
-			priority = driver.PriorityHigh
-		}
-		driver.GetManager().Register(&microphone{id: source.ID()}, driver.Info{
-			Label:      source.ID(),
-			DeviceType: driver.Microphone,
-			Priority:   priority,
-		})
-	}
-}
-
-func (m *microphone) Open() error {
-	var err error
-	m.c, err = pulse.NewClient()
-	if err != nil {
-		return err
-	}
-
-	return nil
-}
-
-func (m *microphone) Close() error {
-	if m.samplesChan != nil {
-		close(m.samplesChan)
-		m.samplesChan = nil
-	}
-
-	m.c.Close()
-	return nil
-}
-
-func (m *microphone) AudioRecord(p prop.Media) (audio.Reader, error) {
-	var options []pulse.RecordOption
-	if p.ChannelCount == 1 {
-		options = append(options, pulse.RecordMono)
-	} else {
-		options = append(options, pulse.RecordStereo)
-	}
-	latency := p.Latency.Seconds()
-
-	src, err := m.c.SourceByID(m.id)
-	if err != nil {
-		return nil, err
-	}
-
-	options = append(options,
-		pulse.RecordSampleRate(p.SampleRate),
-		pulse.RecordLatency(latency),
-		pulse.RecordSource(src),
-	)
-
-	samplesChan := make(chan []int16, 1)
-
-	handler := func(b []int16) (int, error) {
-		samplesChan <- b
-		return len(b), nil
-	}
-
-	stream, err := m.c.NewRecord(pulse.Int16Writer(handler), options...)
-	if err != nil {
-		return nil, err
-	}
-
-	reader := audio.ReaderFunc(func() (wave.Audio, error) {
-		buff, ok := <-samplesChan
-		if !ok {
-			stream.Close()
-			return nil, io.EOF
-		}
-
-		a := wave.NewInt16Interleaved(
-			wave.ChunkInfo{
-				Channels: p.ChannelCount,
-				Len:      len(buff) / p.ChannelCount,
-			},
-		)
-		copy(a.Data, buff)
-
-		return a, nil
-	})
-
-	stream.Start()
-	m.samplesChan = samplesChan
-	return reader, nil
-}
-
-func (m *microphone) Properties() []prop.Media {
-	// TODO: Get actual properties
-	monoProp := prop.Media{
-		Audio: prop.Audio{
-			SampleRate:   48000,
-			Latency:      time.Millisecond * 20,
-			ChannelCount: 1,
-		},
-	}
-
-	stereoProp := monoProp
-	stereoProp.ChannelCount = 2
-
-	return []prop.Media{monoProp, stereoProp}
-}

pkg/driver/microphone/microphone_windows.go

@@ -1,347 +0,0 @@
-package microphone
-
-import (
-	"errors"
-	"golang.org/x/sys/windows"
-	"io"
-	"time"
-	"unsafe"
-
-	"github.com/pion/mediadevices/pkg/driver"
-	"github.com/pion/mediadevices/pkg/io/audio"
-	"github.com/pion/mediadevices/pkg/prop"
-	"github.com/pion/mediadevices/pkg/wave"
-)
-
-const (
-	// bufferNumber * prop.Audio.Latency is the maximum blockable duration
-	// to get data without dropping chunks.
-	bufferNumber = 5
-)
-
-// Windows APIs
-var (
-	winmm                 = windows.NewLazySystemDLL("Winmm.dll")
-	waveInOpen            = winmm.NewProc("waveInOpen")
-	waveInStart           = winmm.NewProc("waveInStart")
-	waveInStop            = winmm.NewProc("waveInStop")
-	waveInReset           = winmm.NewProc("waveInReset")
-	waveInClose           = winmm.NewProc("waveInClose")
-	waveInPrepareHeader   = winmm.NewProc("waveInPrepareHeader")
-	waveInAddBuffer       = winmm.NewProc("waveInAddBuffer")
-	waveInUnprepareHeader = winmm.NewProc("waveInUnprepareHeader")
-)
-
-type buffer struct {
-	waveHdr
-	data []int16
-}
-
-func newBuffer(samples int) *buffer {
-	b := make([]int16, samples)
-	return &buffer{
-		waveHdr: waveHdr{
-			// Sharing Go memory with Windows C API without reference.
-			// Make sure that the lifetime of the buffer struct is longer
-			// than the final access from cbWaveIn.
-			lpData:         uintptr(unsafe.Pointer(&b[0])),
-			dwBufferLength: uint32(samples * 2),
-		},
-		data: b,
-	}
-}
-
-type microphone struct {
-	hWaveIn windows.Pointer
-	buf     map[uintptr]*buffer
-	chBuf   chan *buffer
-	closed  chan struct{}
-}
-
-func init() {
-	// TODO: enum devices
-	driver.GetManager().Register(&microphone{}, driver.Info{
-		Label:      "default",
-		DeviceType: driver.Microphone,
-	})
-}
-
-func (m *microphone) Open() error {
-	m.chBuf = make(chan *buffer)
-	m.buf = make(map[uintptr]*buffer)
-	m.closed = make(chan struct{})
-	return nil
-}
-
-func (m *microphone) cbWaveIn(hWaveIn windows.Pointer, uMsg uint, dwInstance, dwParam1, dwParam2 *int32) uintptr {
-	switch uMsg {
-	case MM_WIM_DATA:
-		b := m.buf[uintptr(unsafe.Pointer(dwParam1))]
-		m.chBuf <- b
-
-	case MM_WIM_OPEN:
-	case MM_WIM_CLOSE:
-		close(m.chBuf)
-	}
-	return 0
-}
-
-func (m *microphone) Close() error {
-	if m.hWaveIn == nil {
-		return nil
-	}
-	close(m.closed)
-
-	ret, _, _ := waveInStop.Call(
-		uintptr(unsafe.Pointer(m.hWaveIn)),
-	)
-	if err := errWinmm[ret]; err != nil {
-		return err
-	}
-	// All enqueued buffers are marked done by waveInReset.
-	ret, _, _ = waveInReset.Call(
-		uintptr(unsafe.Pointer(m.hWaveIn)),
-	)
-	if err := errWinmm[ret]; err != nil {
-		return err
-	}
-	for _, buf := range m.buf {
-		// Detach buffers from waveIn API.
-		ret, _, _ := waveInUnprepareHeader.Call(
-			uintptr(unsafe.Pointer(m.hWaveIn)),
-			uintptr(unsafe.Pointer(&buf.waveHdr)),
-			uintptr(unsafe.Sizeof(buf.waveHdr)),
-		)
-		if err := errWinmm[ret]; err != nil {
-			return err
-		}
-	}
-	// Now, it's ready to free the buffers.
-	// As microphone struct still has reference to the buffers,
-	// they will be GC-ed once microphone is reopened or unreferenced.
-
-	ret, _, _ = waveInClose.Call(
-		uintptr(unsafe.Pointer(m.hWaveIn)),
-	)
-	if err := errWinmm[ret]; err != nil {
-		return err
-	}
-	<-m.chBuf
-	m.hWaveIn = nil
-
-	return nil
-}
-
-func (m *microphone) AudioRecord(p prop.Media) (audio.Reader, error) {
-	for i := 0; i < bufferNumber; i++ {
-		b := newBuffer(
-			int(uint64(p.Latency) * uint64(p.SampleRate) / uint64(time.Second)),
-		)
-		// Map the buffer by its data head address to restore access to the Go struct
-		// in callback function. Don't resize the buffer after it.
-		m.buf[uintptr(unsafe.Pointer(&b.waveHdr))] = b
-	}
-
-	waveFmt := &waveFormatEx{
-		wFormatTag:      WAVE_FORMAT_PCM,
-		nChannels:       uint16(p.ChannelCount),
-		nSamplesPerSec:  uint32(p.SampleRate),
-		nAvgBytesPerSec: uint32(p.SampleRate * p.ChannelCount * 2),
-		nBlockAlign:     uint16(p.ChannelCount * 2),
-		wBitsPerSample:  16,
-	}
-	ret, _, _ := waveInOpen.Call(
-		uintptr(unsafe.Pointer(&m.hWaveIn)),
-		WAVE_MAPPER,
-		uintptr(unsafe.Pointer(waveFmt)),
-		windows.NewCallback(m.cbWaveIn),
-		0,
-		CALLBACK_FUNCTION,
-	)
-	if err := errWinmm[ret]; err != nil {
-		return nil, err
-	}
-
-	for _, buf := range m.buf {
-		// Attach buffers to waveIn API.
-		ret, _, _ := waveInPrepareHeader.Call(
-			uintptr(unsafe.Pointer(m.hWaveIn)),
-			uintptr(unsafe.Pointer(&buf.waveHdr)),
-			uintptr(unsafe.Sizeof(buf.waveHdr)),
-		)
-		if err := errWinmm[ret]; err != nil {
-			return nil, err
-		}
-	}
-	for _, buf := range m.buf {
-		// Enqueue buffers.
-		ret, _, _ := waveInAddBuffer.Call(
-			uintptr(unsafe.Pointer(m.hWaveIn)),
-			uintptr(unsafe.Pointer(&buf.waveHdr)),
-			uintptr(unsafe.Sizeof(buf.waveHdr)),
-		)
-		if err := errWinmm[ret]; err != nil {
-			return nil, err
-		}
-	}
-
-	ret, _, _ = waveInStart.Call(
-		uintptr(unsafe.Pointer(m.hWaveIn)),
-	)
-	if err := errWinmm[ret]; err != nil {
-		return nil, err
-	}
-
-	// TODO: detect microphone device disconnection and return EOF
-
-	reader := audio.ReaderFunc(func() (wave.Audio, error) {
-		b, ok := <-m.chBuf
-		if !ok {
-			return nil, io.EOF
-		}
-
-		select {
-		case <-m.closed:
-		default:
-			// Re-enqueue used buffer.
-			ret, _, _ := waveInAddBuffer.Call(
-				uintptr(unsafe.Pointer(m.hWaveIn)),
-				uintptr(unsafe.Pointer(&b.waveHdr)),
-				uintptr(unsafe.Sizeof(b.waveHdr)),
-			)
-			if err := errWinmm[ret]; err != nil {
-				return nil, err
-			}
-		}
-
-		a := wave.NewInt16Interleaved(
-			wave.ChunkInfo{
-				Channels: p.ChannelCount,
-				Len:      (int(b.waveHdr.dwBytesRecorded) / 2) / p.ChannelCount,
-			},
-		)
-
-		j := 0
-		for i := 0; i < a.Size.Len; i++ {
-			for ch := 0; ch < a.Size.Channels; ch++ {
-				a.SetInt16(i, ch, wave.Int16Sample(b.data[j]))
-				j++
-			}
-		}
-
-		return a, nil
-	})
-	return reader, nil
-}
-
-func (m *microphone) Properties() []prop.Media {
-	// TODO: Get actual properties
-	monoProp := prop.Media{
-		Audio: prop.Audio{
-			SampleRate:   48000,
-			Latency:      time.Millisecond * 20,
-			ChannelCount: 1,
-		},
-	}
-
-	stereoProp := monoProp
-	stereoProp.ChannelCount = 2
-
-	return []prop.Media{monoProp, stereoProp}
-}
-
-// Windows API structures
-
-type waveFormatEx struct {
-	wFormatTag      uint16
-	nChannels       uint16
-	nSamplesPerSec  uint32
-	nAvgBytesPerSec uint32
-	nBlockAlign     uint16
-	wBitsPerSample  uint16
-	cbSize          uint16
-}
-
-type waveHdr struct {
-	lpData          uintptr
-	dwBufferLength  uint32
-	dwBytesRecorded uint32
-	dwUser          *uint32
-	dwFlags         uint32
-	dwLoops         uint32
-	lpNext          *waveHdr
-	reserved        *uint32
-}
-
-// Windows consts
-
-const (
-	MMSYSERR_NOERROR      = 0
-	MMSYSERR_ERROR        = 1
-	MMSYSERR_BADDEVICEID  = 2
-	MMSYSERR_NOTENABLED   = 3
-	MMSYSERR_ALLOCATED    = 4
-	MMSYSERR_INVALHANDLE  = 5
-	MMSYSERR_NODRIVER     = 6
-	MMSYSERR_NOMEM        = 7
-	MMSYSERR_NOTSUPPORTED = 8
-	MMSYSERR_BADERRNUM    = 9
-	MMSYSERR_INVALFLAG    = 10
-	MMSYSERR_INVALPARAM   = 11
-	MMSYSERR_HANDLEBUSY   = 12
-	MMSYSERR_INVALIDALIAS = 13
-	MMSYSERR_BADDB        = 14
-	MMSYSERR_KEYNOTFOUND  = 15
-	MMSYSERR_READERROR    = 16
-	MMSYSERR_WRITEERROR   = 17
-	MMSYSERR_DELETEERROR  = 18
-	MMSYSERR_VALNOTFOUND  = 19
-	MMSYSERR_NODRIVERCB   = 20
-
-	WAVERR_BADFORMAT    = 32
-	WAVERR_STILLPLAYING = 33
-	WAVERR_UNPREPARED   = 34
-	WAVERR_SYNC         = 35
-
-	WAVE_MAPPER     = 0xFFFF
-	WAVE_FORMAT_PCM = 1
-
-	CALLBACK_NULL     = 0
-	CALLBACK_WINDOW   = 0x10000
-	CALLBACK_TASK     = 0x20000
-	CALLBACK_FUNCTION = 0x30000
-	CALLBACK_THREAD   = CALLBACK_TASK
-	CALLBACK_EVENT    = 0x50000
-
-	MM_WIM_OPEN  = 0x3BE
-	MM_WIM_CLOSE = 0x3BF
-	MM_WIM_DATA  = 0x3C0
-)
-
-var errWinmm = map[uintptr]error{
-	MMSYSERR_NOERROR:      nil,
-	MMSYSERR_ERROR:        errors.New("error"),
-	MMSYSERR_BADDEVICEID:  errors.New("bad device id"),
-	MMSYSERR_NOTENABLED:   errors.New("not enabled"),
-	MMSYSERR_ALLOCATED:    errors.New("already allocated"),
-	MMSYSERR_INVALHANDLE:  errors.New("invalid handler"),
-	MMSYSERR_NODRIVER:     errors.New("no driver"),
-	MMSYSERR_NOMEM:        errors.New("no memory"),
-	MMSYSERR_NOTSUPPORTED: errors.New("not supported"),
-	MMSYSERR_BADERRNUM:    errors.New("band error number"),
-	MMSYSERR_INVALFLAG:    errors.New("invalid flag"),
-	MMSYSERR_INVALPARAM:   errors.New("invalid param"),
-	MMSYSERR_HANDLEBUSY:   errors.New("handle busy"),
-	MMSYSERR_INVALIDALIAS: errors.New("invalid alias"),
-	MMSYSERR_BADDB:        errors.New("bad db"),
-	MMSYSERR_KEYNOTFOUND:  errors.New("key not found"),
-	MMSYSERR_READERROR:    errors.New("read error"),
-	MMSYSERR_WRITEERROR:   errors.New("write error"),
-	MMSYSERR_DELETEERROR:  errors.New("delete error"),
-	MMSYSERR_VALNOTFOUND:  errors.New("value not found"),
-	MMSYSERR_NODRIVERCB:   errors.New("no driver cb"),
-	WAVERR_BADFORMAT:      errors.New("bad format"),
-	WAVERR_STILLPLAYING:   errors.New("still playing"),
-	WAVERR_UNPREPARED:     errors.New("unprepared"),
-	WAVERR_SYNC:           errors.New("sync"),
-}

pkg/driver/screen/x11_linux.go

@@ -68,9 +68,9 @@ func (s *screen) VideoRecord(p prop.Media) (video.Reader, error) {
 	var dst image.RGBA
 	reader := s.reader
 
-	r := video.ReaderFunc(func() (image.Image, error) {
+	r := video.ReaderFunc(func() (image.Image, func(), error) {
 		<-s.tick.C
-		return reader.Read().ToRGBA(&dst), nil
+		return reader.Read().ToRGBA(&dst), func() {}, nil
 	})
 	return r, nil
 }

pkg/driver/videotest/dummy.go

@@ -103,10 +103,10 @@ func (d *dummy) VideoRecord(p prop.Media) (video.Reader, error) {
 	d.tick = tick
 	closed := d.closed
 
-	r := video.ReaderFunc(func() (image.Image, error) {
+	r := video.ReaderFunc(func() (image.Image, func(), error) {
 		select {
 		case <-closed:
-			return nil, io.EOF
+			return nil, func() {}, io.EOF
 		default:
 		}
 
@@ -130,7 +130,7 @@ func (d *dummy) VideoRecord(p prop.Media) (video.Reader, error) {
 			CStride:        p.Width / 2,
 			SubsampleRatio: image.YCbCrSubsampleRatio422,
 			Rect:           image.Rect(0, 0, p.Width, p.Height),
-		}, nil
+		}, func() {}, nil
 	})
 
 	return r, nil

pkg/frame/compressed.go

@@ -6,6 +6,7 @@ import (
 	"image/jpeg"
 )
 
-func decodeMJPEG(frame []byte, width, height int) (image.Image, error) {
-	return jpeg.Decode(bytes.NewReader(frame))
+func decodeMJPEG(frame []byte, width, height int) (image.Image, func(), error) {
+	img, err := jpeg.Decode(bytes.NewReader(frame))
+	return img, func() {}, err
 }

pkg/frame/frame.go

@@ -3,12 +3,12 @@ package frame
 import "image"
 
 type Decoder interface {
-	Decode(frame []byte, width, height int) (image.Image, error)
+	Decode(frame []byte, width, height int) (image.Image, func(), error)
 }
 
 // DecoderFunc is a proxy type for Decoder
-type decoderFunc func(frame []byte, width, height int) (image.Image, error)
+type decoderFunc func(frame []byte, width, height int) (image.Image, func(), error)
 
-func (f decoderFunc) Decode(frame []byte, width, height int) (image.Image, error) {
+func (f decoderFunc) Decode(frame []byte, width, height int) (image.Image, func(), error) {
 	return f(frame, width, height)
 }

pkg/frame/yuv.go

@@ -5,13 +5,13 @@ import (
 	"image"
 )
 
-func decodeI420(frame []byte, width, height int) (image.Image, error) {
+func decodeI420(frame []byte, width, height int) (image.Image, func(), error) {
 	yi := width * height
 	cbi := yi + width*height/4
 	cri := cbi + width*height/4
 
 	if cri > len(frame) {
-		return nil, fmt.Errorf("frame length (%d) less than expected (%d)", len(frame), cri)
+		return nil, func() {}, fmt.Errorf("frame length (%d) less than expected (%d)", len(frame), cri)
 	}
 
 	return &image.YCbCr{
@@ -22,15 +22,15 @@ func decodeI420(frame []byte, width, height int) (image.Image, error) {
 		CStride:        width / 2,
 		SubsampleRatio: image.YCbCrSubsampleRatio420,
 		Rect:           image.Rect(0, 0, width, height),
-	}, nil
+	}, func() {}, nil
 }
 
-func decodeNV21(frame []byte, width, height int) (image.Image, error) {
+func decodeNV21(frame []byte, width, height int) (image.Image, func(), error) {
 	yi := width * height
 	ci := yi + width*height/2
 
 	if ci > len(frame) {
-		return nil, fmt.Errorf("frame length (%d) less than expected (%d)", len(frame), ci)
+		return nil, func() {}, fmt.Errorf("frame length (%d) less than expected (%d)", len(frame), ci)
 	}
 
 	var cb, cr []byte
@@ -47,5 +47,5 @@ func decodeNV21(frame []byte, width, height int) (image.Image, error) {
 		CStride:        width / 2,
 		SubsampleRatio: image.YCbCrSubsampleRatio420,
 		Rect:           image.Rect(0, 0, width, height),
-	}, nil
+	}, func() {}, nil
 }

pkg/frame/yuv_cgo.go

@@ -12,13 +12,13 @@ import (
 // void decodeUYVYCGO(uint8_t* y, uint8_t* cb, uint8_t* cr, uint8_t* uyvy, int width, int height);
 import "C"
 
-func decodeYUY2(frame []byte, width, height int) (image.Image, error) {
+func decodeYUY2(frame []byte, width, height int) (image.Image, func(), error) {
 	yi := width * height
 	ci := yi / 2
 	fi := yi + 2*ci
 
 	if len(frame) != fi {
-		return nil, fmt.Errorf("frame length (%d) less than expected (%d)", len(frame), fi)
+		return nil, func() {}, fmt.Errorf("frame length (%d) less than expected (%d)", len(frame), fi)
 	}
 
 	y := make([]byte, yi)
@@ -41,16 +41,16 @@ func decodeYUY2(frame []byte, width, height int) (image.Image, error) {
 		CStride:        width / 2,
 		SubsampleRatio: image.YCbCrSubsampleRatio422,
 		Rect:           image.Rect(0, 0, width, height),
-	}, nil
+	}, func() {}, nil
 }
 
-func decodeUYVY(frame []byte, width, height int) (image.Image, error) {
+func decodeUYVY(frame []byte, width, height int) (image.Image, func(), error) {
 	yi := width * height
 	ci := yi / 2
 	fi := yi + 2*ci
 
 	if len(frame) != fi {
-		return nil, fmt.Errorf("frame length (%d) less than expected (%d)", len(frame), fi)
+		return nil, func() {}, fmt.Errorf("frame length (%d) less than expected (%d)", len(frame), fi)
 	}
 
 	y := make([]byte, yi)
@@ -73,5 +73,5 @@ func decodeUYVY(frame []byte, width, height int) (image.Image, error) {
 		CStride:        width / 2,
 		SubsampleRatio: image.YCbCrSubsampleRatio422,
 		Rect:           image.Rect(0, 0, width, height),
-	}, nil
+	}, func() {}, nil
 }

pkg/frame/yuv_nocgo.go

@@ -7,13 +7,13 @@ import (
 	"image"
 )
 
-func decodeYUY2(frame []byte, width, height int) (image.Image, error) {
+func decodeYUY2(frame []byte, width, height int) (image.Image, func(), error) {
 	yi := width * height
 	ci := yi / 2
 	fi := yi + 2*ci
 
 	if len(frame) != fi {
-		return nil, fmt.Errorf("frame length (%d) less than expected (%d)", len(frame), fi)
+		return nil, func() {}, fmt.Errorf("frame length (%d) less than expected (%d)", len(frame), fi)
 	}
 
 	y := make([]byte, yi)
@@ -39,16 +39,16 @@ func decodeYUY2(frame []byte, width, height int) (image.Image, error) {
 		CStride:        width / 2,
 		SubsampleRatio: image.YCbCrSubsampleRatio422,
 		Rect:           image.Rect(0, 0, width, height),
-	}, nil
+	}, func() {}, nil
 }
 
-func decodeUYVY(frame []byte, width, height int) (image.Image, error) {
+func decodeUYVY(frame []byte, width, height int) (image.Image, func(), error) {
 	yi := width * height
 	ci := yi / 2
 	fi := yi + 2*ci
 
 	if len(frame) != fi {
-		return nil, fmt.Errorf("frame length (%d) less than expected (%d)", len(frame), fi)
+		return nil, func() {}, fmt.Errorf("frame length (%d) less than expected (%d)", len(frame), fi)
 	}
 
 	y := make([]byte, yi)
@@ -74,5 +74,5 @@ func decodeUYVY(frame []byte, width, height int) (image.Image, error) {
 		CStride:        width / 2,
 		SubsampleRatio: image.YCbCrSubsampleRatio422,
 		Rect:           image.Rect(0, 0, width, height),
-	}, nil
+	}, func() {}, nil
 }

pkg/frame/yuv_test.go

@@ -27,7 +27,7 @@ func TestDecodeYUY2(t *testing.T) {
 		Rect:           image.Rect(0, 0, width, height),
 	}
 
-	img, err := decodeYUY2(input, width, height)
+	img, _, err := decodeYUY2(input, width, height)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -56,7 +56,7 @@ func TestDecodeUYVY(t *testing.T) {
 		Rect:           image.Rect(0, 0, width, height),
 	}
 
-	img, err := decodeUYVY(input, width, height)
+	img, _, err := decodeUYVY(input, width, height)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -77,7 +77,7 @@ func BenchmarkDecodeYUY2(b *testing.B) {
 		b.Run(fmt.Sprintf("%dx%d", sz.width, sz.height), func(b *testing.B) {
 			input := make([]byte, sz.width*sz.height*2)
 			for i := 0; i < b.N; i++ {
-				_, err := decodeYUY2(input, sz.width, sz.height)
+				_, _, err := decodeYUY2(input, sz.width, sz.height)
 				if err != nil {
 					b.Fatal(err)
 				}

pkg/io/audio/audio.go

@@ -5,13 +5,14 @@ import (
 )
 
 type Reader interface {
-	Read() (wave.Audio, error)
+	Read() (chunk wave.Audio, release func(), err error)
 }
 
-type ReaderFunc func() (wave.Audio, error)
+type ReaderFunc func() (chunk wave.Audio, release func(), err error)
 
-func (rf ReaderFunc) Read() (wave.Audio, error) {
-	return rf()
+func (rf ReaderFunc) Read() (chunk wave.Audio, release func(), err error) {
+	chunk, release, err = rf()
+	return
 }
 
 // TransformFunc produces a new Reader that will produces a transformed audio

pkg/io/audio/broadcast.go

@@ -0,0 +1,76 @@
+package audio
+
+import (
+	"errors"
+
+	"github.com/pion/mediadevices/pkg/io"
+	"github.com/pion/mediadevices/pkg/wave"
+)
+
+var errEmptySource = errors.New("Source can't be nil")
+
+// Broadcaster is a specialized video broadcaster.
+type Broadcaster struct {
+	ioBroadcaster *io.Broadcaster
+}
+
+type BroadcasterConfig struct {
+	Core *io.BroadcasterConfig
+}
+
+// NewBroadcaster creates a new broadcaster. Source is expected to drop chunks
+// when any of the readers is slower than the source.
+func NewBroadcaster(source Reader, config *BroadcasterConfig) *Broadcaster {
+	var coreConfig *io.BroadcasterConfig
+
+	if config != nil {
+		coreConfig = config.Core
+	}
+
+	broadcaster := io.NewBroadcaster(io.ReaderFunc(func() (interface{}, func(), error) {
+		return source.Read()
+	}), coreConfig)
+
+	return &Broadcaster{broadcaster}
+}
+
+// NewReader creates a new reader. Each reader will retrieve the same data from the source.
+// copyFn is used to copy the data from the source to individual readers. Broadcaster uses a small ring
+// buffer, this means that slow readers might miss some data if they're really late and the data is no longer
+// in the ring buffer.
+func (broadcaster *Broadcaster) NewReader(copyChunk bool) Reader {
+	copyFn := func(src interface{}) interface{} { return src }
+
+	if copyChunk {
+		buffer := wave.NewBuffer()
+		copyFn = func(src interface{}) interface{} {
+			realSrc, _ := src.(wave.Audio)
+			buffer.StoreCopy(realSrc)
+			return buffer.Load()
+		}
+	}
+
+	reader := broadcaster.ioBroadcaster.NewReader(copyFn)
+	return ReaderFunc(func() (wave.Audio, func(), error) {
+		data, _, err := reader.Read()
+		chunk, _ := data.(wave.Audio)
+		return chunk, func() {}, err
+	})
+}
+
+// ReplaceSource replaces the underlying source. This operation is thread safe.
+func (broadcaster *Broadcaster) ReplaceSource(source Reader) error {
+	return broadcaster.ioBroadcaster.ReplaceSource(io.ReaderFunc(func() (interface{}, func(), error) {
+		return source.Read()
+	}))
+}
+
+// Source retrieves the underlying source. This operation is thread safe.
+func (broadcaster *Broadcaster) Source() Reader {
+	source := broadcaster.ioBroadcaster.Source()
+	return ReaderFunc(func() (wave.Audio, func(), error) {
+		data, _, err := source.Read()
+		img, _ := data.(wave.Audio)
+		return img, func() {}, err
+	})
+}

pkg/io/audio/broadcast_test.go

@@ -0,0 +1,54 @@
+package audio
+
+import (
+	"reflect"
+	"testing"
+
+	"github.com/pion/mediadevices/pkg/wave"
+)
+
+func TestBroadcast(t *testing.T) {
+	chunk := wave.NewFloat32Interleaved(wave.ChunkInfo{
+		Len:          8,
+		Channels:     2,
+		SamplingRate: 48000,
+	})
+
+	source := ReaderFunc(func() (wave.Audio, func(), error) {
+		return chunk, func() {}, nil
+	})
+
+	broadcaster := NewBroadcaster(source, nil)
+	readerWithoutCopy1 := broadcaster.NewReader(false)
+	readerWithoutCopy2 := broadcaster.NewReader(false)
+	actualWithoutCopy1, _, err := readerWithoutCopy1.Read()
+	if err != nil {
+		t.Fatal(err)
+	}
+	actualWithoutCopy2, _, err := readerWithoutCopy2.Read()
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if &actualWithoutCopy1.(*wave.Float32Interleaved).Data[0] != &actualWithoutCopy2.(*wave.Float32Interleaved).Data[0] {
+		t.Fatal("Expected underlying buffer for frame with copy to be the same from broadcaster's buffer")
+	}
+
+	if !reflect.DeepEqual(chunk, actualWithoutCopy1) {
+		t.Fatal("Expected actual frame without copy to be the same with the original")
+	}
+
+	readerWithCopy := broadcaster.NewReader(true)
+	actualWithCopy, _, err := readerWithCopy.Read()
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if &actualWithCopy.(*wave.Float32Interleaved).Data[0] == &actualWithoutCopy1.(*wave.Float32Interleaved).Data[0] {
+		t.Fatal("Expected underlying buffer for frame with copy to be different from broadcaster's buffer")
+	}
+
+	if !reflect.DeepEqual(chunk, actualWithCopy) {
+		t.Fatal("Expected actual frame without copy to be the same with the original")
+	}
+}

pkg/io/audio/buffer.go

@@ -13,15 +13,15 @@ func NewBuffer(nSamples int) TransformFunc {
 	var inBuff wave.Audio
 
 	return func(r Reader) Reader {
-		return ReaderFunc(func() (wave.Audio, error) {
+		return ReaderFunc(func() (wave.Audio, func(), error) {
 			for {
 				if inBuff != nil && inBuff.ChunkInfo().Len >= nSamples {
 					break
 				}
 
-				buff, err := r.Read()
+				buff, _, err := r.Read()
 				if err != nil {
-					return nil, err
+					return nil, func() {}, err
 				}
 				switch b := buff.(type) {
 				case *wave.Float32Interleaved:
@@ -59,7 +59,7 @@ func NewBuffer(nSamples int) TransformFunc {
 					ib.Size.Len += b.Size.Len
 
 				default:
-					return nil, errUnsupported
+					return nil, func() {}, errUnsupported
 				}
 			}
 			switch ib := inBuff.(type) {
@@ -71,7 +71,7 @@ func NewBuffer(nSamples int) TransformFunc {
 				copy(ibCopy.Data, ib.Data)
 				ib.Data = ib.Data[n:]
 				ib.Size.Len -= nSamples
-				return &ibCopy, nil
+				return &ibCopy, func() {}, nil
 
 			case *wave.Float32Interleaved:
 				ibCopy := *ib
@@ -81,9 +81,9 @@ func NewBuffer(nSamples int) TransformFunc {
 				copy(ibCopy.Data, ib.Data)
 				ib.Data = ib.Data[n:]
 				ib.Size.Len -= nSamples
-				return &ibCopy, nil
+				return &ibCopy, func() {}, nil
 			}
-			return nil, errUnsupported
+			return nil, func() {}, errUnsupported
 		})
 	}
 }

pkg/io/audio/buffer_test.go

@@ -49,16 +49,16 @@ func TestBuffer(t *testing.T) {
 	trans := NewBuffer(3)
 
 	var iSent int
-	r := trans(ReaderFunc(func() (wave.Audio, error) {
+	r := trans(ReaderFunc(func() (wave.Audio, func(), error) {
 		if iSent < len(input) {
 			iSent++
-			return input[iSent-1], nil
+			return input[iSent-1], func() {}, nil
 		}
-		return nil, io.EOF
+		return nil, func() {}, io.EOF
 	}))
 
 	for i := 0; ; i++ {
-		a, err := r.Read()
+		a, _, err := r.Read()
 		if err != nil {
 			if err == io.EOF && i >= len(expected) {
 				break

pkg/io/audio/detect.go

@@ -0,0 +1,55 @@
+package audio
+
+import (
+	"time"
+
+	"github.com/pion/mediadevices/pkg/prop"
+	"github.com/pion/mediadevices/pkg/wave"
+)
+
+// DetectChanges will detect chunk and audio property changes. For audio property detection,
+// since it's time related, interval will be used to determine the sample rate.
+func DetectChanges(interval time.Duration, onChange func(prop.Media)) TransformFunc {
+	return func(r Reader) Reader {
+		var currentProp prop.Media
+		var chunkCount uint
+		return ReaderFunc(func() (wave.Audio, func(), error) {
+			var dirty bool
+
+			chunk, _, err := r.Read()
+			if err != nil {
+				return nil, func() {}, err
+			}
+
+			info := chunk.ChunkInfo()
+			if currentProp.ChannelCount != info.Channels {
+				currentProp.ChannelCount = info.Channels
+				dirty = true
+			}
+
+			if currentProp.SampleRate != info.SamplingRate {
+				currentProp.SampleRate = info.SamplingRate
+				dirty = true
+			}
+
+			var latency time.Duration
+			if currentProp.SampleRate != 0 {
+				latency = time.Duration(chunk.ChunkInfo().Len) * time.Second / time.Nanosecond / time.Duration(currentProp.SampleRate)
+			}
+			if currentProp.Latency != latency {
+				currentProp.Latency = latency
+				dirty = true
+			}
+
+			// TODO: Also detect sample format changes?
+			// TODO: Add audio detect changes. As of now, there's no useful property to track.
+
+			if dirty {
+				onChange(currentProp)
+			}
+
+			chunkCount++
+			return chunk, func() {}, nil
+		})
+	}
+}

pkg/io/audio/detect_test.go

@@ -0,0 +1,76 @@
+package audio
+
+import (
+	"reflect"
+	"testing"
+	"time"
+
+	"github.com/pion/mediadevices/pkg/prop"
+	"github.com/pion/mediadevices/pkg/wave"
+)
+
+func TestDetectChanges(t *testing.T) {
+	buildSource := func(p prop.Media) (Reader, func(prop.Media)) {
+		return ReaderFunc(func() (wave.Audio, func(), error) {
+				return wave.NewFloat32Interleaved(wave.ChunkInfo{
+					Len:          960,
+					Channels:     p.ChannelCount,
+					SamplingRate: p.SampleRate,
+				}), func() {}, nil
+			}), func(newProp prop.Media) {
+				p = newProp
+			}
+	}
+
+	t.Run("OnChangeCalledBeforeFirstFrame", func(t *testing.T) {
+		var detectBeforeFirstChunk bool
+		var expected prop.Media
+		var actual prop.Media
+		expected.ChannelCount = 2
+		expected.SampleRate = 48000
+		expected.Latency = time.Millisecond * 20
+		src, _ := buildSource(expected)
+		src = DetectChanges(time.Second, func(p prop.Media) {
+			actual = p
+			detectBeforeFirstChunk = true
+		})(src)
+
+		_, _, err := src.Read()
+		if err != nil {
+			t.Fatal(err)
+		}
+
+		if !detectBeforeFirstChunk {
+			t.Fatal("on change callback should have called before first chunk")
+		}
+
+		if !reflect.DeepEqual(actual, expected) {
+			t.Fatalf("Received an unexpected prop\nExpected:\n%v\nActual:\n%v\n", expected, actual)
+		}
+	})
+
+	t.Run("DetectChangesOnEveryUpdate", func(t *testing.T) {
+		var expected prop.Media
+		var actual prop.Media
+		expected.ChannelCount = 2
+		expected.SampleRate = 48000
+		expected.Latency = 20 * time.Millisecond
+		src, update := buildSource(expected)
+		src = DetectChanges(time.Second, func(p prop.Media) {
+			actual = p
+		})(src)
+
+		for channelCount := 1; channelCount < 8; channelCount++ {
+			expected.ChannelCount = channelCount
+			update(expected)
+			_, _, err := src.Read()
+			if err != nil {
+				t.Fatal(err)
+			}
+
+			if !reflect.DeepEqual(actual, expected) {
+				t.Fatalf("Received an unexpected prop\nExpected:\n%v\nActual:\n%v\n", expected, actual)
+			}
+		}
+	})
+}

pkg/io/audio/mixer.go

@@ -8,14 +8,14 @@ import (
 // NewChannelMixer creates audio transform to mix audio channels.
 func NewChannelMixer(channels int, mixer mixer.ChannelMixer) TransformFunc {
 	return func(r Reader) Reader {
-		return ReaderFunc(func() (wave.Audio, error) {
-			buff, err := r.Read()
+		return ReaderFunc(func() (wave.Audio, func(), error) {
+			buff, _, err := r.Read()
 			if err != nil {
-				return nil, err
+				return nil, func() {}, err
 			}
 			ci := buff.ChunkInfo()
 			if ci.Channels == channels {
-				return buff, nil
+				return buff, func() {}, nil
 			}
 
 			ci.Channels = channels
@@ -32,9 +32,9 @@ func NewChannelMixer(channels int, mixer mixer.ChannelMixer) TransformFunc {
 				mixed = wave.NewFloat32NonInterleaved(ci)
 			}
 			if err := mixer.Mix(mixed, buff); err != nil {
-				return nil, err
+				return nil, func() {}, err
 			}
-			return mixed, nil
+			return mixed, func() {}, nil
 		})
 	}
 }

pkg/io/audio/mixer_test.go

@@ -34,16 +34,16 @@ func TestMixer(t *testing.T) {
 	trans := NewChannelMixer(1, &mixer.MonoMixer{})
 
 	var iSent int
-	r := trans(ReaderFunc(func() (wave.Audio, error) {
+	r := trans(ReaderFunc(func() (wave.Audio, func(), error) {
 		if iSent < len(input) {
 			iSent++
-			return input[iSent-1], nil
+			return input[iSent-1], func() {}, nil
 		}
-		return nil, io.EOF
+		return nil, func() {}, io.EOF
 	}))
 
 	for i := 0; ; i++ {
-		a, err := r.Read()
+		a, _, err := r.Read()
 		if err != nil {
 			if err == io.EOF && i >= len(expected) {
 				break

pkg/io/broadcast.go

@@ -0,0 +1,162 @@
+package io
+
+import (
+	"fmt"
+	"sync/atomic"
+	"time"
+)
+
+const (
+	maskReading                = 1 << 63
+	defaultBroadcasterRingSize = 32
+	// TODO: If the data source has fps greater than 30, they'll see some
+	// 			 fps fluctuation. But, 30 fps should be enough for general cases.
+	defaultBroadcasterRingPollDuration = time.Millisecond * 33
+)
+
+var errEmptySource = fmt.Errorf("Source can't be nil")
+
+type broadcasterData struct {
+	data  interface{}
+	count uint32
+	err   error
+}
+
+type broadcasterRing struct {
+	// reading (1 bit) + reserved (31 bits) + data count (32 bits)
+	// IMPORTANT: state has to be the first element in struct, otherwise LoadUint64 will panic in 32 bits systems
+	//            due to unallignment
+	state        uint64
+	buffer       []atomic.Value
+	pollDuration time.Duration
+}
+
+func newBroadcasterRing(size uint, pollDuration time.Duration) *broadcasterRing {
+	return &broadcasterRing{buffer: make([]atomic.Value, size), pollDuration: pollDuration}
+}
+
+func (ring *broadcasterRing) index(count uint32) int {
+	return int(count) % len(ring.buffer)
+}
+
+func (ring *broadcasterRing) acquire(count uint32) func(*broadcasterData) {
+	// Reader has reached the latest data, should read from the source.
+	// Only allow 1 reader to read from the source. When there are more than 1 readers,
+	// the other readers will need to share the same data that the first reader gets from
+	// the source.
+	state := uint64(count)
+	if atomic.CompareAndSwapUint64(&ring.state, state, state|maskReading) {
+		return func(data *broadcasterData) {
+			i := ring.index(count)
+			ring.buffer[i].Store(data)
+			atomic.StoreUint64(&ring.state, uint64(count+1))
+		}
+	}
+
+	return nil
+}
+
+func (ring *broadcasterRing) get(count uint32) *broadcasterData {
+	for {
+		reading := uint64(count) | maskReading
+		// TODO: since it's lockless, it spends a lot of resources in the scheduling.
+		for atomic.LoadUint64(&ring.state) == reading {
+			// Yield current goroutine to let other goroutines to run instead
+			time.Sleep(ring.pollDuration)
+		}
+
+		i := ring.index(count)
+		data := ring.buffer[i].Load().(*broadcasterData)
+		if data.count == count {
+			return data
+		}
+
+		count++
+	}
+}
+
+func (ring *broadcasterRing) lastCount() uint32 {
+	// ring.state always keeps track the next count, so we need to subtract it by 1 to get the
+	// last count
+	return uint32(atomic.LoadUint64(&ring.state)) - 1
+}
+
+// Broadcaster is a generic pull-based broadcaster. Broadcaster is unique in a sense that
+// readers can come and go at anytime, and readers don't need to close or notify broadcaster.
+type Broadcaster struct {
+	source atomic.Value
+	buffer *broadcasterRing
+}
+
+// BroadcasterConfig is a config to control broadcaster behaviour
+type BroadcasterConfig struct {
+	// BufferSize configures the underlying ring buffer size that's being used
+	// to avoid data lost for late readers. The default value is 32.
+	BufferSize uint
+	// PollDuration configures the sleep duration in waiting for new data to come.
+	// The default value is 33 ms.
+	PollDuration time.Duration
+}
+
+// NewBroadcaster creates a new broadcaster. Source is expected to drop frames
+// when any of the readers is slower than the source.
+func NewBroadcaster(source Reader, config *BroadcasterConfig) *Broadcaster {
+	pollDuration := defaultBroadcasterRingPollDuration
+	var bufferSize uint = defaultBroadcasterRingSize
+	if config != nil {
+		if config.PollDuration != 0 {
+			pollDuration = config.PollDuration
+		}
+
+		if config.BufferSize != 0 {
+			bufferSize = config.BufferSize
+		}
+	}
+
+	var broadcaster Broadcaster
+	broadcaster.buffer = newBroadcasterRing(bufferSize, pollDuration)
+	broadcaster.ReplaceSource(source)
+
+	return &broadcaster
+}
+
+// NewReader creates a new reader. Each reader will retrieve the same data from the source.
+// copyFn is used to copy the data from the source to individual readers. Broadcaster uses a small ring
+// buffer, this means that slow readers might miss some data if they're really late and the data is no longer
+// in the ring buffer.
+func (broadcaster *Broadcaster) NewReader(copyFn func(interface{}) interface{}) Reader {
+	currentCount := broadcaster.buffer.lastCount()
+
+	return ReaderFunc(func() (data interface{}, release func(), err error) {
+		currentCount++
+		if push := broadcaster.buffer.acquire(currentCount); push != nil {
+			data, _, err = broadcaster.source.Load().(Reader).Read()
+			push(&broadcasterData{
+				data:  data,
+				err:   err,
+				count: currentCount,
+			})
+		} else {
+			ringData := broadcaster.buffer.get(currentCount)
+			data, err, currentCount = ringData.data, ringData.err, ringData.count
+		}
+
+		data = copyFn(data)
+		return
+	})
+}
+
+// ReplaceSource replaces the underlying source. This operation is thread safe.
+func (broadcaster *Broadcaster) ReplaceSource(source Reader) error {
+	if source == nil {
+		return errEmptySource
+	}
+
+	broadcaster.source.Store(source)
+	return nil
+}
+
+// ReplaceSource retrieves the underlying source. This operation is thread safe.
+func (broadcaster *Broadcaster) Source() Reader {
+	return broadcaster.source.Load().(Reader)
+}

pkg/io/broadcast_test.go

@@ -0,0 +1,148 @@
+package io
+
+import (
+	"fmt"
+	"runtime"
+	"sync"
+	"sync/atomic"
+	"testing"
+	"time"
+)
+
+func TestBroadcast(t *testing.T) {
+	// https://github.com/pion/mediadevices/issues/198
+	if runtime.GOOS == "darwin" {
+		t.Skip("Skipping because Darwin CI is not reliable for timing related tests.")
+	}
+	frames := make([]int, 5*30) // 5 seconds worth of frames
+	for i := range frames {
+		frames[i] = i
+	}
+
+	routinePauseConds := []struct {
+		src          bool
+		dst          bool
+		expectedFPS  float64
+		expectedDrop float64
+	}{
+		{
+			src:         false,
+			dst:         false,
+			expectedFPS: 30,
+		},
+		{
+			src:          true,
+			dst:          false,
+			expectedFPS:  20,
+			expectedDrop: 10,
+		},
+		{
+			src:          false,
+			dst:          true,
+			expectedFPS:  20,
+			expectedDrop: 10,
+		},
+	}
+
+	for _, pauseCond := range routinePauseConds {
+		pauseCond := pauseCond
+		t.Run(fmt.Sprintf("SrcPause-%v/DstPause-%v", pauseCond.src, pauseCond.dst), func(t *testing.T) {
+			for n := 1; n <= 256; n *= 16 {
+				n := n
+
+				t.Run(fmt.Sprintf("Readers-%d", n), func(t *testing.T) {
+					var src Reader
+					interval := time.NewTicker(time.Millisecond * 33) // 30 fps
+					defer interval.Stop()
+					frameCount := 0
+					frameSent := 0
+					lastSend := time.Now()
+					src = ReaderFunc(func() (interface{}, func(), error) {
+						if pauseCond.src && frameSent == 30 {
+							time.Sleep(time.Second)
+						}
+						<-interval.C
+
+						now := time.Now()
+						if interval := now.Sub(lastSend); interval > time.Millisecond*33*3/2 {
+							// Source reader should drop frames to catch up the latest frame.
+							drop := int(interval/(time.Millisecond*33)) - 1
+							frameCount += drop
+							t.Logf("Skipped %d frames", drop)
+						}
+						lastSend = now
+						frame := frames[frameCount]
+						frameCount++
+						frameSent++
+						return frame, func() {}, nil
+					})
+					broadcaster := NewBroadcaster(src, nil)
+					var done uint32
+					duration := time.Second * 3
+					fpsChan := make(chan []float64)
+
+					var wg sync.WaitGroup
+					wg.Add(n)
+					for i := 0; i < n; i++ {
+						go func() {
+							reader := broadcaster.NewReader(func(src interface{}) interface{} { return src })
+							count := 0
+							lastFrameCount := -1
+							droppedFrames := 0
+							wg.Done()
+							wg.Wait()
+							for atomic.LoadUint32(&done) == 0 {
+								if pauseCond.dst && count == 30 {
+									time.Sleep(time.Second)
+								}
+								frame, _, err := reader.Read()
+								if err != nil {
+									t.Error(err)
+								}
+								frameCount := frame.(int)
+								droppedFrames += (frameCount - lastFrameCount - 1)
+								lastFrameCount = frameCount
+								count++
+							}
+
+							fps := float64(count) / duration.Seconds()
+							if fps < pauseCond.expectedFPS-2 || fps > pauseCond.expectedFPS+2 {
+								t.Fatal("Unexpected average FPS")
+							}
+
+							droppedFramesPerSecond := float64(droppedFrames) / duration.Seconds()
+							if droppedFramesPerSecond < pauseCond.expectedDrop-2 || droppedFramesPerSecond > pauseCond.expectedDrop+2 {
+								t.Fatal("Unexpected drop count")
+							}
+
+							fpsChan <- []float64{fps, droppedFramesPerSecond, float64(lastFrameCount)}
+						}()
+					}
+
+					time.Sleep(duration)
+					atomic.StoreUint32(&done, 1)
+
+					var fpsAvg float64
+					var droppedFramesPerSecondAvg float64
+					var lastFrameCountAvg float64
+					var count int
+					for metric := range fpsChan {
+						fps, droppedFramesPerSecond, lastFrameCount := metric[0], metric[1], metric[2]
+						fpsAvg += fps
+						droppedFramesPerSecondAvg += droppedFramesPerSecond
+						lastFrameCountAvg += lastFrameCount
+						count++
+						if count == n {
+							break
+						}
+					}
+
+					t.Log("Average FPS                      :", fpsAvg/float64(n))
+					t.Log("Average dropped frames per second:", droppedFramesPerSecondAvg/float64(n))
+					t.Log("Last frame count (src)           :", frameCount)
+					t.Log("Average last frame count (dst)   :", lastFrameCountAvg/float64(n))
+				})
+			}
+		})
+	}
+}

pkg/io/io.go

@@ -1,11 +0,0 @@
-package io
-
-// Copy copies data from src to dst. If dst is not big enough, return an
-// InsufficientBufferError.
-func Copy(dst, src []byte) (n int, err error) {
-	if len(dst) < len(src) {
-		return 0, &InsufficientBufferError{len(src)}
-	}
-
-	return copy(dst, src), nil
-}

pkg/io/io_test.go

@@ -1,45 +0,0 @@
-package io
-
-import (
-	"log"
-	"testing"
-)
-
-func TestCopy(t *testing.T) {
-	var dst []byte
-	src := make([]byte, 4)
-
-	n, err := Copy(dst, src)
-	if err == nil {
-		t.Fatal("expected err to be non-nill")
-	}
-
-	if n != 0 {
-		t.Fatalf("expected n to be 0, but got %d", n)
-	}
-
-	e, ok := err.(*InsufficientBufferError)
-	if !ok {
-		t.Fatalf("expected error to be InsufficientBufferError")
-	}
-
-	if e.RequiredSize != len(src) {
-		t.Fatalf("expected required size to be %d, but got %d", len(src), e.RequiredSize)
-	}
-
-	dst = make([]byte, 2*e.RequiredSize)
-	n, err = Copy(dst, src)
-	if err != nil {
-		t.Fatalf("expected to not get an error after expanding the buffer")
-	}
-
-	if n != len(src) {
-		t.Fatalf("expected n to be %d, but got %d", len(src), n)
-	}
-
-	for i := 0; i < len(src); i++ {
-		if src[i] != dst[i] {
-			log.Fatalf("expected value at %d to be %d, but got %d", i, src[i], dst[i])
-		}
-	}
-}

pkg/io/reader.go

@@ -0,0 +1,15 @@
+package io
+
+// Reader is a generic data reader. In the future, interface{} should be replaced by a generic type
+// to provide strong type.
+type Reader interface {
+	Read() (data interface{}, release func(), err error)
+}
+
+// ReaderFunc is a proxy type for Reader
+type ReaderFunc func() (data interface{}, release func(), err error)
+
+func (f ReaderFunc) Read() (data interface{}, release func(), err error) {
+	data, release, err = f()
+	return
+}

pkg/io/video/broadcast.go

@@ -0,0 +1,76 @@
+package video
+
+import (
+	"fmt"
+	"image"
+
+	"github.com/pion/mediadevices/pkg/io"
+)
+
+var errEmptySource = fmt.Errorf("Source can't be nil")
+
+// Broadcaster is a specialized video broadcaster.
+type Broadcaster struct {
+	ioBroadcaster *io.Broadcaster
+}
+
+type BroadcasterConfig struct {
+	Core *io.BroadcasterConfig
+}
+
+// NewBroadcaster creates a new broadcaster. Source is expected to drop frames
+// when any of the readers is slower than the source.
+func NewBroadcaster(source Reader, config *BroadcasterConfig) *Broadcaster {
+	var coreConfig *io.BroadcasterConfig
+
+	if config != nil {
+		coreConfig = config.Core
+	}
+
+	broadcaster := io.NewBroadcaster(io.ReaderFunc(func() (interface{}, func(), error) {
+		return source.Read()
+	}), coreConfig)
+
+	return &Broadcaster{broadcaster}
+}
+
+// NewReader creates a new reader. Each reader will retrieve the same data from the source.
+// copyFn is used to copy the data from the source to individual readers. Broadcaster uses a small ring
+// buffer, this means that slow readers might miss some data if they're really late and the data is no longer
+// in the ring buffer.
+func (broadcaster *Broadcaster) NewReader(copyFrame bool) Reader {
+	copyFn := func(src interface{}) interface{} { return src }
+
+	if copyFrame {
+		buffer := NewFrameBuffer(0)
+		copyFn = func(src interface{}) interface{} {
+			realSrc, _ := src.(image.Image)
+			buffer.StoreCopy(realSrc)
+			return buffer.Load()
+		}
+	}
+
+	reader := broadcaster.ioBroadcaster.NewReader(copyFn)
+	return ReaderFunc(func() (image.Image, func(), error) {
+		data, _, err := reader.Read()
+		img, _ := data.(image.Image)
+		return img, func() {}, err
+	})
+}
+
+// ReplaceSource replaces the underlying source. This operation is thread safe.
+func (broadcaster *Broadcaster) ReplaceSource(source Reader) error {
+	return broadcaster.ioBroadcaster.ReplaceSource(io.ReaderFunc(func() (interface{}, func(), error) {
+		return source.Read()
+	}))
+}
+
+// Source retrieves the underlying source. This operation is thread safe.
+func (broadcaster *Broadcaster) Source() Reader {
+	source := broadcaster.ioBroadcaster.Source()
+	return ReaderFunc(func() (image.Image, func(), error) {
+		data, _, err := source.Read()
+		img, _ := data.(image.Image)
+		return img, func() {}, err
+	})
+}

pkg/io/video/broadcast_test.go

@@ -0,0 +1,49 @@
+package video
+
+import (
+	"image"
+	"reflect"
+	"testing"
+)
+
+func TestBroadcast(t *testing.T) {
+	resolution := image.Rect(0, 0, 1920, 1080)
+	img := image.NewGray(resolution)
+	source := ReaderFunc(func() (image.Image, func(), error) {
+		return img, func() {}, nil
+	})
+
+	broadcaster := NewBroadcaster(source, nil)
+	readerWithoutCopy1 := broadcaster.NewReader(false)
+	readerWithoutCopy2 := broadcaster.NewReader(false)
+	actualWithoutCopy1, _, err := readerWithoutCopy1.Read()
+	if err != nil {
+		t.Fatal(err)
+	}
+	actualWithoutCopy2, _, err := readerWithoutCopy2.Read()
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if &actualWithoutCopy1.(*image.Gray).Pix[0] != &actualWithoutCopy2.(*image.Gray).Pix[0] {
+		t.Fatal("Expected underlying buffer for frame with copy to be the same from broadcaster's buffer")
+	}
+
+	if !reflect.DeepEqual(img, actualWithoutCopy1) {
+		t.Fatal("Expected actual frame without copy to be the same with the original")
+	}
+
+	readerWithCopy := broadcaster.NewReader(true)
+	actualWithCopy, _, err := readerWithCopy.Read()
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if &actualWithCopy.(*image.Gray).Pix[0] == &actualWithoutCopy1.(*image.Gray).Pix[0] {
+		t.Fatal("Expected underlying buffer for frame with copy to be different from broadcaster's buffer")
+	}
+
+	if !reflect.DeepEqual(img, actualWithCopy) {
+		t.Fatal("Expected actual frame without copy to be the same with the original")
+	}
+}

pkg/io/video/convert.go

@@ -63,10 +63,10 @@ func imageToYCbCr(dst *image.YCbCr, src image.Image) {
 // ToI420 converts r to a new reader that will output images in I420 format
 func ToI420(r Reader) Reader {
 	var yuvImg image.YCbCr
-	return ReaderFunc(func() (image.Image, error) {
-		img, err := r.Read()
+	return ReaderFunc(func() (image.Image, func(), error) {
+		img, _, err := r.Read()
 		if err != nil {
-			return nil, err
+			return nil, func() {}, err
 		}
 
 		imageToYCbCr(&yuvImg, img)
@@ -79,11 +79,11 @@ func ToI420(r Reader) Reader {
 			i422ToI420(&yuvImg)
 		case image.YCbCrSubsampleRatio420:
 		default:
-			return nil, fmt.Errorf("unsupported pixel format: %s", yuvImg.SubsampleRatio)
+			return nil, func() {}, fmt.Errorf("unsupported pixel format: %s", yuvImg.SubsampleRatio)
 		}
 
 		yuvImg.SubsampleRatio = image.YCbCrSubsampleRatio420
-		return &yuvImg, nil
+		return &yuvImg, func() {}, nil
 	})
 }
 
@@ -130,13 +130,13 @@ func imageToRGBA(dst *image.RGBA, src image.Image) {
 // ToRGBA converts r to a new reader that will output images in RGBA format
 func ToRGBA(r Reader) Reader {
 	var dst image.RGBA
-	return ReaderFunc(func() (image.Image, error) {
-		img, err := r.Read()
+	return ReaderFunc(func() (image.Image, func(), error) {
+		img, _, err := r.Read()
 		if err != nil {
-			return nil, err
+			return nil, func() {}, err
 		}
 
 		imageToRGBA(&dst, img)
-		return &dst, nil
+		return &dst, func() {}, nil
 	})
 }

pkg/io/video/convert_test.go

@@ -144,10 +144,10 @@ func TestToI420(t *testing.T) {
 	for name, c := range cases {
 		c := c
 		t.Run(name, func(t *testing.T) {
-			r := ToI420(ReaderFunc(func() (image.Image, error) {
-				return c.src, nil
+			r := ToI420(ReaderFunc(func() (image.Image, func(), error) {
+				return c.src, func() {}, nil
 			}))
-			out, err := r.Read()
+			out, _, err := r.Read()
 			if err != nil {
 				t.Fatalf("Unexpected error: %v", err)
 			}
@@ -199,10 +199,10 @@ func TestToRGBA(t *testing.T) {
 	for name, c := range cases {
 		c := c
 		t.Run(name, func(t *testing.T) {
-			r := ToRGBA(ReaderFunc(func() (image.Image, error) {
-				return c.src, nil
+			r := ToRGBA(ReaderFunc(func() (image.Image, func(), error) {
+				return c.src, func() {}, nil
 			}))
-			out, err := r.Read()
+			out, _, err := r.Read()
 			if err != nil {
 				t.Fatalf("Unexpected error: %v", err)
 			}
@@ -225,12 +225,12 @@ func BenchmarkToI420(b *testing.B) {
 			for name, img := range cases {
 				img := img
 				b.Run(name, func(b *testing.B) {
-					r := ToI420(ReaderFunc(func() (image.Image, error) {
-						return img, nil
+					r := ToI420(ReaderFunc(func() (image.Image, func(), error) {
+						return img, func() {}, nil
 					}))
 
 					for i := 0; i < b.N; i++ {
-						_, err := r.Read()
+						_, _, err := r.Read()
 						if err != nil {
 							b.Fatalf("Unexpected error: %v", err)
 						}
@@ -253,12 +253,12 @@ func BenchmarkToRGBA(b *testing.B) {
 			for name, img := range cases {
 				img := img
 				b.Run(name, func(b *testing.B) {
-					r := ToRGBA(ReaderFunc(func() (image.Image, error) {
-						return img, nil
+					r := ToRGBA(ReaderFunc(func() (image.Image, func(), error) {
+						return img, func() {}, nil
 					}))
 
 					for i := 0; i < b.N; i++ {
-						_, err := r.Read()
+						_, _, err := r.Read()
 						if err != nil {
 							b.Fatalf("Unexpected error: %v", err)
 						}

pkg/io/video/detect.go

@@ -14,12 +14,12 @@ func DetectChanges(interval time.Duration, onChange func(prop.Media)) TransformF
 		var currentProp prop.Media
 		var lastTaken time.Time
 		var frames uint
-		return ReaderFunc(func() (image.Image, error) {
+		return ReaderFunc(func() (image.Image, func(), error) {
 			var dirty bool
 
-			img, err := r.Read()
+			img, _, err := r.Read()
 			if err != nil {
-				return nil, err
+				return nil, func() {}, err
 			}
 
 			bounds := img.Bounds()
@@ -52,7 +52,7 @@ func DetectChanges(interval time.Duration, onChange func(prop.Media)) TransformF
 			}
 
 			frames++
-			return img, nil
+			return img, func() {}, nil
 		})
 	}
 }

pkg/io/video/detect_test.go

@@ -12,8 +12,8 @@ import (
 
 func BenchmarkDetectChanges(b *testing.B) {
 	var src Reader
-	src = ReaderFunc(func() (image.Image, error) {
-		return image.NewRGBA(image.Rect(0, 0, 1920, 1080)), nil
+	src = ReaderFunc(func() (image.Image, func(), error) {
+		return image.NewRGBA(image.Rect(0, 0, 1920, 1080)), func() {}, nil
 	})
 
 	b.Run("WithoutDetectChanges", func(b *testing.B) {
@@ -40,8 +40,8 @@ func BenchmarkDetectChanges(b *testing.B) {
 
 func TestDetectChanges(t *testing.T) {
 	buildSource := func(p prop.Media) (Reader, func(prop.Media)) {
-		return ReaderFunc(func() (image.Image, error) {
-				return image.NewRGBA(image.Rect(0, 0, p.Width, p.Height)), nil
+		return ReaderFunc(func() (image.Image, func(), error) {
+				return image.NewRGBA(image.Rect(0, 0, p.Width, p.Height)), func() {}, nil
 			}), func(newProp prop.Media) {
 				p = newProp
 			}
@@ -86,7 +86,7 @@ func TestDetectChanges(t *testing.T) {
 			detectBeforeFirstFrame = true
 		})(src)
 
-		frame, err := src.Read()
+		frame, _, err := src.Read()
 		if err != nil {
 			t.Fatal(err)
 		}
@@ -113,7 +113,7 @@ func TestDetectChanges(t *testing.T) {
 				expected.Width = width
 				expected.Height = height
 				update(expected)
-				frame, err := src.Read()
+				frame, _, err := src.Read()
 				if err != nil {
 					t.Fatal(err)
 				}
@@ -143,7 +143,7 @@ func TestDetectChanges(t *testing.T) {
 		})(src)
 
 		for count < 3 {
-			frame, err := src.Read()
+			frame, _, err := src.Read()
 			if err != nil {
 				t.Fatal(err)
 			}

pkg/io/video/scale.go

@@ -156,10 +156,10 @@ func Scale(width, height int, scaler Scaler) TransformFunc {
 			}
 		}
 
-		return ReaderFunc(func() (image.Image, error) {
-			img, err := r.Read()
+		return ReaderFunc(func() (image.Image, func(), error) {
+			img, _, err := r.Read()
 			if err != nil {
-				return nil, err
+				return nil, func() {}, err
 			}
 
 			switch v := img.(type) {
@@ -169,7 +169,7 @@ func Scale(width, height int, scaler Scaler) TransformFunc {
 				scalerCached.Scale(dst, rect, v, v.Rect, draw.Src, nil)
 
 				cloned := *dst // clone metadata
-				return &cloned, nil
+				return &cloned, func() {}, nil
 
 			case *image.YCbCr:
 				ycbcrRealloc(v)
@@ -184,10 +184,10 @@ func Scale(width, height int, scaler Scaler) TransformFunc {
 				scalerCached.Scale(dst, dst.Bounds(), src, src.Bounds(), draw.Src, nil)
 
 				cloned := *(imgScaled.(*image.YCbCr)) // clone metadata
-				return &cloned, nil
+				return &cloned, func() {}, nil
 
 			default:
-				return nil, errUnsupportedImageType
+				return nil, func() {}, errUnsupportedImageType
 			}
 		})
 	}

pkg/io/video/scale_test.go

@@ -215,11 +215,11 @@ func TestScale(t *testing.T) {
 				c := c
 				t.Run(name, func(t *testing.T) {
 					trans := Scale(c.width, c.height, algo)
-					r := trans(ReaderFunc(func() (image.Image, error) {
-						return c.src, nil
+					r := trans(ReaderFunc(func() (image.Image, func(), error) {
+						return c.src, func() {}, nil
 					}))
 					for i := 0; i < 4; i++ {
-						out, err := r.Read()
+						out, _, err := r.Read()
 						if err != nil {
 							t.Fatalf("Unexpected error: %v", err)
 						}
@@ -261,12 +261,12 @@ func BenchmarkScale(b *testing.B) {
 						img := img
 						b.Run(name, func(b *testing.B) {
 							trans := Scale(640, 360, algo)
-							r := trans(ReaderFunc(func() (image.Image, error) {
-								return img, nil
+							r := trans(ReaderFunc(func() (image.Image, func(), error) {
+								return img, func() {}, nil
 							}))
 
 							for i := 0; i < b.N; i++ {
-								_, err := r.Read()
+								_, _, err := r.Read()
 								if err != nil {
 									b.Fatalf("Unexpected error: %v", err)
 								}

pkg/io/video/throttle.go

@@ -10,16 +10,16 @@ import (
 func Throttle(rate float32) TransformFunc {
 	return func(r Reader) Reader {
 		ticker := time.NewTicker(time.Duration(int64(float64(time.Second) / float64(rate))))
-		return ReaderFunc(func() (image.Image, error) {
+		return ReaderFunc(func() (image.Image, func(), error) {
 			for {
-				img, err := r.Read()
+				img, _, err := r.Read()
 				if err != nil {
 					ticker.Stop()
-					return nil, err
+					return nil, func() {}, err
 				}
 				select {
 				case <-ticker.C:
-					return img, nil
+					return img, func() {}, nil
 				default:
 				}
 			}

pkg/io/video/throttle_test.go

@@ -19,14 +19,14 @@ func TestThrottle(t *testing.T) {
 
 	var cntPush int
 	trans := Throttle(50)
-	r := trans(ReaderFunc(func() (image.Image, error) {
+	r := trans(ReaderFunc(func() (image.Image, func(), error) {
 		<-ticker.C
 		cntPush++
-		return img, nil
+		return img, func() {}, nil
 	}))
 
 	for i := 0; i < 20; i++ {
-		_, err := r.Read()
+		_, _, err := r.Read()
 		if err != nil {
 			t.Fatalf("Unexpected error: %v", err)
 		}

pkg/io/video/video.go

@@ -5,13 +5,14 @@ import (
 )
 
 type Reader interface {
-	Read() (img image.Image, err error)
+	Read() (img image.Image, release func(), err error)
 }
 
-type ReaderFunc func() (img image.Image, err error)
+type ReaderFunc func() (img image.Image, release func(), err error)
 
-func (rf ReaderFunc) Read() (img image.Image, err error) {
-	return rf()
+func (rf ReaderFunc) Read() (img image.Image, release func(), err error) {
+	img, release, err = rf()
+	return
 }
 
 // TransformFunc produces a new Reader that will produces a transformed video

pkg/prop/prop.go

@@ -42,10 +42,17 @@ func prettifyStruct(i interface{}) string {
 			value := obj.Field(i)
 
 			padding := strings.Repeat("  ", level)
-			if value.Kind() == reflect.Struct {
+			switch value.Kind() {
+			case reflect.Struct:
 				rows = append(rows, fmt.Sprintf("%s%v:", padding, field.Name))
 				addRows(level+1, value)
-			} else {
+			case reflect.Chan, reflect.Func, reflect.Interface, reflect.Map, reflect.Ptr, reflect.Slice:
+				if value.IsNil() {
+					rows = append(rows, fmt.Sprintf("%s%v: any", padding, field.Name))
+				} else {
+					rows = append(rows, fmt.Sprintf("%s%v: %v", padding, field.Name, value))
+				}
+			default:
 				rows = append(rows, fmt.Sprintf("%s%v: %v", padding, field.Name, value))
 			}
 		}

pkg/wave/buffer.go

@@ -0,0 +1,149 @@
+package wave
+
+import "fmt"
+
+var (
+	errUnsupportedFormat = fmt.Errorf("Unsupported format")
+)
+
+// Buffer is a buffer that can store any audio format.
+type Buffer struct {
+	// TODO: Probably standardize the audio formats so that we don't need to have the following different types
+	//       and duplicated codes for each type
+	bufferFloat32Interleaved    []float32
+	bufferFloat32NonInterleaved [][]float32
+	bufferInt16Interleaved      []int16
+	bufferInt16NonInterleaved   [][]int16
+	tmp                         Audio
+}
+
+// NewBuffer creates a new Buffer instance
+func NewBuffer() *Buffer {
+	return &Buffer{}
+}
+
+// Load loads the current owned Audio
+func (buff *Buffer) Load() Audio {
+	return buff.tmp
+}
+
+// StoreCopy makes a copy of src and store its copy. StoreCopy will reuse as much memory as it can
+// from the previous copies. For example, if StoreCopy is given an audio that has the format from the previous call,
+// StoreCopy will not allocate extra memory and only copy the content from src to the previous buffer.
+func (buff *Buffer) StoreCopy(src Audio) {
+	switch src := src.(type) {
+	case *Float32Interleaved:
+		clone, ok := buff.tmp.(*Float32Interleaved)
+		if ok {
+			*clone = *src
+		} else {
+			copied := *src
+			clone = &copied
+		}
+
+		neededSize := len(src.Data)
+		if len(buff.bufferFloat32Interleaved) < neededSize {
+			if cap(buff.bufferFloat32Interleaved) >= neededSize {
+				buff.bufferFloat32Interleaved = buff.bufferFloat32Interleaved[:neededSize]
+			} else {
+				buff.bufferFloat32Interleaved = make([]float32, neededSize)
+			}
+		}
+
+		copy(buff.bufferFloat32Interleaved, src.Data)
+		clone.Data = buff.bufferFloat32Interleaved
+		buff.tmp = clone
+
+	case *Float32NonInterleaved:
+		clone, ok := buff.tmp.(*Float32NonInterleaved)
+		if ok {
+			*clone = *src
+		} else {
+			copied := *src
+			clone = &copied
+		}
+
+		neededSize := len(src.Data)
+		if len(buff.bufferFloat32NonInterleaved) < neededSize {
+			if cap(buff.bufferFloat32NonInterleaved) >= neededSize {
+				buff.bufferFloat32NonInterleaved = buff.bufferFloat32NonInterleaved[:neededSize]
+			} else {
+				buff.bufferFloat32NonInterleaved = make([][]float32, neededSize)
+			}
+		}
+
+		for i := range src.Data {
+			neededSize := len(src.Data[i])
+			if len(buff.bufferFloat32NonInterleaved[i]) < neededSize {
+				if cap(buff.bufferFloat32NonInterleaved[i]) >= neededSize {
+					buff.bufferFloat32NonInterleaved[i] = buff.bufferFloat32NonInterleaved[i][:neededSize]
+				} else {
+					buff.bufferFloat32NonInterleaved[i] = make([]float32, neededSize)
+				}
+			}
+
+			copy(buff.bufferFloat32NonInterleaved[i], src.Data[i])
+		}
+		clone.Data = buff.bufferFloat32NonInterleaved
+		buff.tmp = clone
+
+	case *Int16Interleaved:
+		clone, ok := buff.tmp.(*Int16Interleaved)
+		if ok {
+			*clone = *src
+		} else {
+			copied := *src
+			clone = &copied
+		}
+
+		neededSize := len(src.Data)
+		if len(buff.bufferInt16Interleaved) < neededSize {
+			if cap(buff.bufferInt16Interleaved) >= neededSize {
+				buff.bufferInt16Interleaved = buff.bufferInt16Interleaved[:neededSize]
+			} else {
+				buff.bufferInt16Interleaved = make([]int16, neededSize)
+			}
+		}
+
+		copy(buff.bufferInt16Interleaved, src.Data)
+		clone.Data = buff.bufferInt16Interleaved
+		buff.tmp = clone
+
+	case *Int16NonInterleaved:
+		clone, ok := buff.tmp.(*Int16NonInterleaved)
+		if ok {
+			*clone = *src
+		} else {
+			copied := *src
+			clone = &copied
+		}
+
+		neededSize := len(src.Data)
+		if len(buff.bufferInt16NonInterleaved) < neededSize {
+			if cap(buff.bufferInt16NonInterleaved) >= neededSize {
+				buff.bufferInt16NonInterleaved = buff.bufferInt16NonInterleaved[:neededSize]
+			} else {
+				buff.bufferInt16NonInterleaved = make([][]int16, neededSize)
+			}
+		}
+
+		for i := range src.Data {
+			neededSize := len(src.Data[i])
+			if len(buff.bufferInt16NonInterleaved[i]) < neededSize {
+				if cap(buff.bufferInt16NonInterleaved[i]) >= neededSize {
+					buff.bufferInt16NonInterleaved[i] = buff.bufferInt16NonInterleaved[i][:neededSize]
+				} else {
+					buff.bufferInt16NonInterleaved[i] = make([]int16, neededSize)
+				}
+			}
+
+			copy(buff.bufferInt16NonInterleaved[i], src.Data[i])
+		}
+		clone.Data = buff.bufferInt16NonInterleaved
+		buff.tmp = clone
+
+	default:
+		// TODO: Should have a routine to convert any format to one of the supported formats above
+		panic(errUnsupportedFormat)
+	}
+}

pkg/wave/buffer_test.go

@@ -0,0 +1,129 @@
+package wave
+
+import (
+	"errors"
+	"fmt"
+	"reflect"
+	"testing"
+)
+
+var (
+	errIdenticalAddress = errors.New("Cloned audio has the same memory address with the original audio")
+)
+
+func TestBufferStoreCopyAndLoad(t *testing.T) {
+	chunkInfo := ChunkInfo{
+		Len:          4,
+		Channels:     2,
+		SamplingRate: 48000,
+	}
+	testCases := map[string]struct {
+		New      func() EditableAudio
+		Update   func(EditableAudio)
+		Validate func(*testing.T, Audio, Audio)
+	}{
+		"Float32Interleaved": {
+			New: func() EditableAudio {
+				return NewFloat32Interleaved(chunkInfo)
+			},
+			Update: func(src EditableAudio) {
+				src.Set(0, 0, Float32Sample(1))
+			},
+			Validate: func(t *testing.T, original Audio, clone Audio) {
+				ok := reflect.ValueOf(original.(*Float32Interleaved).Data).Pointer() != reflect.ValueOf(clone.(*Float32Interleaved).Data).Pointer()
+				if !ok {
+					t.Error(errIdenticalAddress)
+				}
+			},
+		},
+		"Float32NonInterleaved": {
+			New: func() EditableAudio {
+				return NewFloat32NonInterleaved(chunkInfo)
+			},
+			Update: func(src EditableAudio) {
+				src.Set(0, 0, Float32Sample(1))
+			},
+			Validate: func(t *testing.T, original Audio, clone Audio) {
+				originalReal := original.(*Float32NonInterleaved)
+				cloneReal := clone.(*Float32NonInterleaved)
+				if reflect.ValueOf(originalReal.Data).Pointer() == reflect.ValueOf(cloneReal.Data).Pointer() {
+					t.Error(errIdenticalAddress)
+				}
+
+				for i := range cloneReal.Data {
+					if reflect.ValueOf(originalReal.Data[i]).Pointer() == reflect.ValueOf(cloneReal.Data[i]).Pointer() {
+						err := fmt.Errorf("Channel %d memory address should be different", i)
+						t.Errorf("%v: %w", errIdenticalAddress, err)
+					}
+				}
+			},
+		},
+		"Int16Interleaved": {
+			New: func() EditableAudio {
+				return NewInt16Interleaved(chunkInfo)
+			},
+			Update: func(src EditableAudio) {
+				src.Set(1, 1, Int16Sample(2))
+			},
+			Validate: func(t *testing.T, original Audio, clone Audio) {
+				ok := reflect.ValueOf(original.(*Int16Interleaved).Data).Pointer() != reflect.ValueOf(clone.(*Int16Interleaved).Data).Pointer()
+				if !ok {
+					t.Error(errIdenticalAddress)
+				}
+			},
+		},
+		"Int16NonInterleaved": {
+			New: func() EditableAudio {
+				return NewInt16NonInterleaved(chunkInfo)
+			},
+			Update: func(src EditableAudio) {
+				src.Set(1, 1, Int16Sample(2))
+			},
+			Validate: func(t *testing.T, original Audio, clone Audio) {
+				originalReal := original.(*Int16NonInterleaved)
+				cloneReal := clone.(*Int16NonInterleaved)
+				if reflect.ValueOf(originalReal.Data).Pointer() == reflect.ValueOf(cloneReal.Data).Pointer() {
+					t.Error(errIdenticalAddress)
+				}
+
+				for i := range cloneReal.Data {
+					if reflect.ValueOf(originalReal.Data[i]).Pointer() == reflect.ValueOf(cloneReal.Data[i]).Pointer() {
+						err := fmt.Errorf("Channel %d memory address should be different", i)
+						t.Errorf("%v: %w", errIdenticalAddress, err)
+					}
+				}
+			},
+		},
+	}
+
+	buffer := NewBuffer()
+
+	for name, testCase := range testCases {
+		// Since the test also wants to make sure that Copier can convert from 1 type to another,
+		// t.Run is not ideal since it'll run the tests separately
+		t.Log("Testing", name)
+
+		src := testCase.New()
+		src.Set(0, 0, Int16Sample(1))
+		buffer.StoreCopy(src)
+
+		testCase.Validate(t, src, buffer.Load())
+
+		if !reflect.DeepEqual(buffer.Load(), src) {
+			t.Fatalf(`Expected the copied audio chunk to be identical with the source
+
+Expected:
+%v
+
+Actual:
+%v
+			`, src, buffer.Load())
+		}
+
+		testCase.Update(src)
+		buffer.StoreCopy(src)
+		if !reflect.DeepEqual(buffer.Load(), src) {
+			t.Fatal("Expected the copied audio chunk to be identical with the source after an update in source")
+		}
+	}
+}

sampler.go

@@ -4,6 +4,7 @@ import (
 	"math"
 	"time"
 
+	"github.com/pion/webrtc/v2"
 	"github.com/pion/webrtc/v2/pkg/media"
 )
 
@@ -11,7 +12,7 @@ type samplerFunc func(b []byte) error
 
 // newVideoSampler creates a video sampler that uses the actual video frame rate and
 // the codec's clock rate to come up with a duration for each sample.
-func newVideoSampler(t LocalTrack) samplerFunc {
+func newVideoSampler(t *webrtc.Track) samplerFunc {
 	clockRate := float64(t.Codec().ClockRate)
 	lastTimestamp := time.Now()
 
@@ -27,7 +28,7 @@ func newVideoSampler(t LocalTrack) samplerFunc {
 
 // newAudioSampler creates a audio sampler that uses a fixed latency and
 // the codec's clock rate to come up with a duration for each sample.
-func newAudioSampler(t LocalTrack, latency time.Duration) samplerFunc {
+func newAudioSampler(t *webrtc.Track, latency time.Duration) samplerFunc {
 	samples := uint32(math.Round(float64(t.Codec().ClockRate) * latency.Seconds()))
 	return samplerFunc(func(b []byte) error {
 		return t.WriteSample(media.Sample{Data: b, Samples: samples})

source.go

@@ -0,0 +1 @@
+package mediadevices

track.go

@@ -2,237 +2,324 @@ package mediadevices
 
 import (
 	"errors"
+	"image"
 	"math/rand"
 	"sync"
 
 	"github.com/pion/mediadevices/pkg/codec"
 	"github.com/pion/mediadevices/pkg/driver"
-	mio "github.com/pion/mediadevices/pkg/io"
+	"github.com/pion/mediadevices/pkg/io/audio"
+	"github.com/pion/mediadevices/pkg/io/video"
+	"github.com/pion/mediadevices/pkg/wave"
 	"github.com/pion/webrtc/v2"
-	"github.com/pion/webrtc/v2/pkg/media"
 )
 
-// Tracker is an interface that represent MediaStreamTrack
+var (
+	errInvalidDriverType      = errors.New("invalid driver type")
+	errNotFoundPeerConnection = errors.New("failed to find given peer connection")
+)
+
+// Source is a generic representation of a media source
+type Source interface {
+	ID() string
+	Close() error
+}
+
+// VideoSource is a specific type of media source that emits a series of video frames
+type VideoSource interface {
+	video.Reader
+	Source
+}
+
+// AudioSource is a specific type of media source that emits a series of audio chunks
+type AudioSource interface {
+	audio.Reader
+	Source
+}
+
+// Track is an interface that represent MediaStreamTrack
 // Reference: https://w3c.github.io/mediacapture-main/#mediastreamtrack
-type Tracker interface {
-	Track() *webrtc.Track
-	LocalTrack() LocalTrack
-	Stop()
+type Track interface {
+	Source
 	// OnEnded registers a handler to receive an error from the media stream track.
 	// If the error is already occured before registering, the handler will be
 	// immediately called.
 	OnEnded(func(error))
+	Kind() MediaDeviceType
+	// Bind binds the current track source to the given peer connection. In Pion/webrtc v3, the bind
+	// call will happen automatically after the SDP negotiation. Users won't need to call this manually.
+	Bind(*webrtc.PeerConnection) (*webrtc.Track, error)
+	// Unbind is the clean up operation that should be called after Bind. Similar to Bind, unbind will
+	// be called automatically in the future.
+	Unbind(*webrtc.PeerConnection) error
 }
 
-type LocalTrack interface {
-	WriteSample(s media.Sample) error
-	Codec() *webrtc.RTPCodec
-	ID() string
-	Kind() webrtc.RTPCodecType
+type baseTrack struct {
+	Source
+	err                   error
+	onErrorHandler        func(error)
+	mu                    sync.Mutex
+	endOnce               sync.Once
+	kind                  MediaDeviceType
+	selector              *CodecSelector
+	activePeerConnections map[*webrtc.PeerConnection]chan<- chan<- struct{}
 }
 
-type track struct {
-	localTrack LocalTrack
-	d          driver.Driver
-	sample     samplerFunc
-	encoder    codec.ReadCloser
-
-	onErrorHandler func(error)
-	err            error
-	mu             sync.Mutex
-	endOnce        sync.Once
+func newBaseTrack(source Source, kind MediaDeviceType, selector *CodecSelector) *baseTrack {
+	return &baseTrack{
+		Source:                source,
+		kind:                  kind,
+		selector:              selector,
+		activePeerConnections: make(map[*webrtc.PeerConnection]chan<- chan<- struct{}),
+	}
 }
 
-func newTrack(opts *MediaDevicesOptions, d driver.Driver, constraints MediaTrackConstraints) (*track, error) {
-	var encoderBuilders []encoderBuilder
-	var rtpCodecs []*webrtc.RTPCodec
-	var buildSampler func(t LocalTrack) samplerFunc
-	var err error
-
-	err = d.Open()
-	if err != nil {
-		return nil, err
-	}
-
-	switch r := d.(type) {
-	case driver.VideoRecorder:
-		rtpCodecs = opts.codecs[webrtc.RTPCodecTypeVideo]
-		buildSampler = newVideoSampler
-		encoderBuilders, err = newVideoEncoderBuilders(r, constraints)
-	case driver.AudioRecorder:
-		rtpCodecs = opts.codecs[webrtc.RTPCodecTypeAudio]
-		buildSampler = func(t LocalTrack) samplerFunc {
-			return newAudioSampler(t, constraints.selectedMedia.Latency)
-		}
-		encoderBuilders, err = newAudioEncoderBuilders(r, constraints)
-	default:
-		err = errors.New("newTrack: invalid driver type")
-	}
-
-	if err != nil {
-		d.Close()
-		return nil, err
-	}
-
-	for _, builder := range encoderBuilders {
-		var matchedRTPCodec *webrtc.RTPCodec
-		for _, rtpCodec := range rtpCodecs {
-			if rtpCodec.Name == builder.name {
-				matchedRTPCodec = rtpCodec
-				break
-			}
-		}
-
-		if matchedRTPCodec == nil {
-			continue
-		}
-
-		localTrack, err := opts.trackGenerator(
-			matchedRTPCodec.PayloadType,
-			rand.Uint32(),
-			d.ID(),
-			matchedRTPCodec.Type.String(),
-			matchedRTPCodec,
-		)
-		if err != nil {
-			continue
-		}
-
-		encoder, err := builder.build()
-		if err != nil {
-			continue
-		}
-
-		t := track{
-			localTrack: localTrack,
-			sample:     buildSampler(localTrack),
-			d:          d,
-			encoder:    encoder,
-		}
-		go t.start()
-		return &t, nil
-	}
-
-	d.Close()
-	return nil, errors.New("newTrack: failed to find a matching codec")
+// Kind returns track's kind
+func (track *baseTrack) Kind() MediaDeviceType {
+	return track.kind
 }
 
 // OnEnded sets an error handler. When a track has been created and started, if an
 // error occurs, handler will get called with the error given to the parameter.
-func (t *track) OnEnded(handler func(error)) {
-	t.mu.Lock()
-	t.onErrorHandler = handler
-	err := t.err
-	t.mu.Unlock()
+func (track *baseTrack) OnEnded(handler func(error)) {
+	track.mu.Lock()
+	track.onErrorHandler = handler
+	err := track.err
+	track.mu.Unlock()
 
 	if err != nil && handler != nil {
 		// Already errored.
-		t.endOnce.Do(func() {
+		track.endOnce.Do(func() {
 			handler(err)
 		})
 	}
 }
 
 // onError is a callback when an error occurs
-func (t *track) onError(err error) {
-	t.mu.Lock()
-	t.err = err
-	handler := t.onErrorHandler
-	t.mu.Unlock()
+func (track *baseTrack) onError(err error) {
+	track.mu.Lock()
+	track.err = err
+	handler := track.onErrorHandler
+	track.mu.Unlock()
 
 	if handler != nil {
-		t.endOnce.Do(func() {
+		track.endOnce.Do(func() {
 			handler(err)
 		})
 	}
 }
 
-// start starts the data flow from the driver all the way to the localTrack
-func (t *track) start() {
-	var n int
-	var err error
-	buff := make([]byte, 1024)
-	for {
-		n, err = t.encoder.Read(buff)
-		if err != nil {
-			if e, ok := err.(*mio.InsufficientBufferError); ok {
-				buff = make([]byte, 2*e.RequiredSize)
-				continue
+func (track *baseTrack) bind(pc *webrtc.PeerConnection, encodedReader codec.ReadCloser, selectedCodec *codec.RTPCodec, sampler func(*webrtc.Track) samplerFunc) (*webrtc.Track, error) {
+	track.mu.Lock()
+	defer track.mu.Unlock()
+
+	webrtcTrack, err := pc.NewTrack(selectedCodec.PayloadType, rand.Uint32(), track.ID(), selectedCodec.MimeType)
+	if err != nil {
+		return nil, err
+	}
+
+	sample := sampler(webrtcTrack)
+	signalCh := make(chan chan<- struct{})
+	track.activePeerConnections[pc] = signalCh
+
+	go func() {
+		var doneCh chan<- struct{}
+		defer func() {
+			encodedReader.Close()
+
+			// When there's another call to unbind, it won't block since we mark the signalCh to be closed
+			close(signalCh)
+			if doneCh != nil {
+				close(doneCh)
+			}
+		}()
+
+		for {
+			select {
+			case doneCh = <-signalCh:
+				return
+			default:
 			}
 
-			t.onError(err)
-			return
-		}
+			buff, _, err := encodedReader.Read()
+			if err != nil {
+				track.onError(err)
+				return
+			}
 
-		if err := t.sample(buff[:n]); err != nil {
-			t.onError(err)
-			return
+			if err := sample(buff); err != nil {
+				track.onError(err)
+				return
+			}
 		}
+	}()
+
+	return webrtcTrack, nil
+}
+
+func (track *baseTrack) unbind(pc *webrtc.PeerConnection) error {
+	track.mu.Lock()
+	defer track.mu.Unlock()
+
+	ch, ok := track.activePeerConnections[pc]
+	if !ok {
+		return errNotFoundPeerConnection
+	}
+
+	doneCh := make(chan struct{})
+	ch <- doneCh
+	<-doneCh
+	delete(track.activePeerConnections, pc)
+	return nil
+}
+
+func newTrackFromDriver(d driver.Driver, constraints MediaTrackConstraints, selector *CodecSelector) (Track, error) {
+	if err := d.Open(); err != nil {
+		return nil, err
+	}
+
+	switch recorder := d.(type) {
+	case driver.VideoRecorder:
+		return newVideoTrackFromDriver(d, recorder, constraints, selector)
+	case driver.AudioRecorder:
+		return newAudioTrackFromDriver(d, recorder, constraints, selector)
+	default:
+		panic(errInvalidDriverType)
 	}
 }
 
-// Stop stops the underlying driver and encoder
-func (t *track) Stop() {
-	t.d.Close()
-	t.encoder.Close()
+// VideoTrack is a specific track type that contains video source which allows multiple readers to access, and manipulate.
+type VideoTrack struct {
+	*baseTrack
+	*video.Broadcaster
 }
 
-func (t *track) Track() *webrtc.Track {
-	return t.localTrack.(*webrtc.Track)
+// NewVideoTrack constructs a new VideoTrack
+func NewVideoTrack(source VideoSource, selector *CodecSelector) Track {
+	return newVideoTrackFromReader(source, source, selector)
 }
 
-func (t *track) LocalTrack() LocalTrack {
-	return t.localTrack
+func newVideoTrackFromReader(source Source, reader video.Reader, selector *CodecSelector) Track {
+	base := newBaseTrack(source, VideoInput, selector)
+	wrappedReader := video.ReaderFunc(func() (img image.Image, release func(), err error) {
+		img, _, err = reader.Read()
+		if err != nil {
+			base.onError(err)
+		}
+		return img, func() {}, err
+	})
+
+	// TODO: Allow users to configure broadcaster
+	broadcaster := video.NewBroadcaster(wrappedReader, nil)
+
+	return &VideoTrack{
+		baseTrack:   base,
+		Broadcaster: broadcaster,
+	}
 }
 
-// encoderBuilder is a generic encoder builder that acts as a delegator for codec.VideoEncoderBuilder and
-// codec.AudioEncoderBuilder. The idea of having a delegator is to reduce redundant codes that are being
-// duplicated for managing video and audio.
-type encoderBuilder struct {
-	name  string
-	build func() (codec.ReadCloser, error)
-}
-
-// newVideoEncoderBuilders transforms video given by VideoRecorder with the video transformer that is passed through
-// constraints and create a list of generic encoder builders
-func newVideoEncoderBuilders(vr driver.VideoRecorder, constraints MediaTrackConstraints) ([]encoderBuilder, error) {
-	r, err := vr.VideoRecord(constraints.selectedMedia)
+// newVideoTrackFromDriver is an internal video track creation from driver
+func newVideoTrackFromDriver(d driver.Driver, recorder driver.VideoRecorder, constraints MediaTrackConstraints, selector *CodecSelector) (Track, error) {
+	reader, err := recorder.VideoRecord(constraints.selectedMedia)
 	if err != nil {
 		return nil, err
 	}
 
-	if constraints.VideoTransform != nil {
-		r = constraints.VideoTransform(r)
-	}
-
-	encoderBuilders := make([]encoderBuilder, len(constraints.VideoEncoderBuilders))
-	for i, b := range constraints.VideoEncoderBuilders {
-		encoderBuilders[i].name = b.Name()
-		encoderBuilders[i].build = func() (codec.ReadCloser, error) {
-			return b.BuildVideoEncoder(r, constraints.selectedMedia)
-		}
-	}
-	return encoderBuilders, nil
+	return newVideoTrackFromReader(d, reader, selector), nil
 }
 
-// newAudioEncoderBuilders transforms audio given by AudioRecorder with the audio transformer that is passed through
-// constraints and create a list of generic encoder builders
-func newAudioEncoderBuilders(ar driver.AudioRecorder, constraints MediaTrackConstraints) ([]encoderBuilder, error) {
-	r, err := ar.AudioRecord(constraints.selectedMedia)
+// Transform transforms the underlying source by applying the given fns in serial order
+func (track *VideoTrack) Transform(fns ...video.TransformFunc) {
+	src := track.Broadcaster.Source()
+	track.Broadcaster.ReplaceSource(video.Merge(fns...)(src))
+}
+
+func (track *VideoTrack) Bind(pc *webrtc.PeerConnection) (*webrtc.Track, error) {
+	reader := track.NewReader(false)
+	inputProp, err := detectCurrentVideoProp(track.Broadcaster)
 	if err != nil {
 		return nil, err
 	}
 
-	if constraints.AudioTransform != nil {
-		r = constraints.AudioTransform(r)
+	wantCodecs := pc.GetRegisteredRTPCodecs(webrtc.RTPCodecTypeVideo)
+	encodedReader, selectedCodec, err := track.selector.selectVideoCodec(wantCodecs, reader, inputProp)
+	if err != nil {
+		return nil, err
 	}
 
-	encoderBuilders := make([]encoderBuilder, len(constraints.AudioEncoderBuilders))
-	for i, b := range constraints.AudioEncoderBuilders {
-		encoderBuilders[i].name = b.Name()
-		encoderBuilders[i].build = func() (codec.ReadCloser, error) {
-			return b.BuildAudioEncoder(r, constraints.selectedMedia)
-		}
-	}
-	return encoderBuilders, nil
+	return track.bind(pc, encodedReader, selectedCodec, newVideoSampler)
+}
+
+func (track *VideoTrack) Unbind(pc *webrtc.PeerConnection) error {
+	return track.unbind(pc)
+}
+
+// AudioTrack is a specific track type that contains audio source which allows multiple readers to access, and
+// manipulate.
+type AudioTrack struct {
+	*baseTrack
+	*audio.Broadcaster
+}
+
+// NewAudioTrack constructs a new VideoTrack
+func NewAudioTrack(source AudioSource, selector *CodecSelector) Track {
+	return newAudioTrackFromReader(source, source, selector)
+}
+
+func newAudioTrackFromReader(source Source, reader audio.Reader, selector *CodecSelector) Track {
+	base := newBaseTrack(source, AudioInput, selector)
+	wrappedReader := audio.ReaderFunc(func() (chunk wave.Audio, release func(), err error) {
+		chunk, _, err = reader.Read()
+		if err != nil {
+			base.onError(err)
+		}
+		return chunk, func() {}, err
+	})
+
+	// TODO: Allow users to configure broadcaster
+	broadcaster := audio.NewBroadcaster(wrappedReader, nil)
+
+	return &AudioTrack{
+		baseTrack:   base,
+		Broadcaster: broadcaster,
+	}
+}
+
+// newAudioTrackFromDriver is an internal audio track creation from driver
+func newAudioTrackFromDriver(d driver.Driver, recorder driver.AudioRecorder, constraints MediaTrackConstraints, selector *CodecSelector) (Track, error) {
+	reader, err := recorder.AudioRecord(constraints.selectedMedia)
+	if err != nil {
+		return nil, err
+	}
+
+	// FIXME: The current audio detection and audio encoder can only work with a static latency. Since the latency from the driver
+	//        can fluctuate, we need to stabilize it. Maybe there's a better way for doing this?
+	reader = audio.NewBuffer(int(constraints.selectedMedia.Latency.Seconds() * float64(constraints.selectedMedia.SampleRate)))(reader)
+	return newAudioTrackFromReader(d, reader, selector), nil
+}
+
+// Transform transforms the underlying source by applying the given fns in serial order
+func (track *AudioTrack) Transform(fns ...audio.TransformFunc) {
+	src := track.Broadcaster.Source()
+	track.Broadcaster.ReplaceSource(audio.Merge(fns...)(src))
+}
+
+func (track *AudioTrack) Bind(pc *webrtc.PeerConnection) (*webrtc.Track, error) {
+	reader := track.NewReader(false)
+	inputProp, err := detectCurrentAudioProp(track.Broadcaster)
+	if err != nil {
+		return nil, err
+	}
+
+	wantCodecs := pc.GetRegisteredRTPCodecs(webrtc.RTPCodecTypeAudio)
+	encodedReader, selectedCodec, err := track.selector.selectAudioCodec(wantCodecs, reader, inputProp)
+	if err != nil {
+		return nil, err
+	}
+
+	return track.bind(pc, encodedReader, selectedCodec, func(t *webrtc.Track) samplerFunc { return newAudioSampler(t, inputProp.Latency) })
+}
+
+func (track *AudioTrack) Unbind(pc *webrtc.PeerConnection) error {
+	return track.unbind(pc)
 }

track_test.go

@@ -10,7 +10,7 @@ func TestOnEnded(t *testing.T) {
 	errExpected := errors.New("an error")
 
 	t.Run("ErrorAfterRegister", func(t *testing.T) {
-		tr := &track{}
+		tr := &baseTrack{}
 
 		called := make(chan error, 1)
 		tr.OnEnded(func(error) {
@@ -35,7 +35,7 @@ func TestOnEnded(t *testing.T) {
 	})
 
 	t.Run("ErrorBeforeRegister", func(t *testing.T) {
-		tr := &track{}
+		tr := &baseTrack{}
 
 		tr.onError(errExpected)