Default high profile for x264 when possible

miniaudio supports various backends. This will help reducing code
surface

examples/webrtc/main.go

@@ -9,14 +9,14 @@ import (
 	"github.com/pion/mediadevices/pkg/prop"
 	"github.com/pion/webrtc/v2"
 
-	// If you don't like vpx, you can also use x264 by importing as below
+	// If you don't like x264, you can also use vpx by importing as below
-	// "github.com/pion/mediadevices/pkg/codec/x264" // This is required to use h264 video encoder
+	// "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"
 	// 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 VP8/VP9 video encoder
+	"github.com/pion/mediadevices/pkg/codec/opus" // This is required to use opus audio encoder
-	"github.com/pion/mediadevices/pkg/codec/vpx"  // This is required to use VP8/VP9 video 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.
@@ -44,18 +44,18 @@ func main() {
 	signal.Decode(signal.MustReadStdin(), &offer)
 
 	// Create a new RTCPeerConnection
-	vp8Params, err := vpx.NewVP8Params()
+	x264Params, err := x264.NewParams()
 	if err != nil {
 		panic(err)
 	}
-	vp8Params.BitRate = 300_000 // 300kbps
+	x264Params.BitRate = 500_000 // 500kbps
 
 	opusParams, err := opus.NewParams()
 	if err != nil {
 		panic(err)
 	}
 	codecSelector := mediadevices.NewCodecSelector(
-		mediadevices.WithVideoEncoders(&vp8Params),
+		mediadevices.WithVideoEncoders(&x264Params),
 		mediadevices.WithAudioEncoders(&opusParams),
 	)
 

mediadevices_bench_test.go

@@ -1,3 +1,5 @@
+// +build e2e
+
 package mediadevices
 
 import (

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/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

@@ -1,138 +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, func(), error) {
-		buff, ok := <-samplesChan
-		if !ok {
-			stream.Close()
-			return nil, func() {}, io.EOF
-		}
-
-		a := wave.NewInt16Interleaved(
-			wave.ChunkInfo{
-				Channels:     p.ChannelCount,
-				Len:          len(buff) / p.ChannelCount,
-				SamplingRate: p.SampleRate,
-			},
-		)
-		copy(a.Data, buff)
-
-		return a, func() {}, 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

@@ -1,348 +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, func(), error) {
-		b, ok := <-m.chBuf
-		if !ok {
-			return nil, func() {}, 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, func() {}, err
-			}
-		}
-
-		a := wave.NewInt16Interleaved(
-			wave.ChunkInfo{
-				Channels:     p.ChannelCount,
-				Len:          (int(b.waveHdr.dwBytesRecorded) / 2) / p.ChannelCount,
-				SamplingRate: p.SampleRate,
-			},
-		)
-
-		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, func() {}, 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"),
-}

track.go

@@ -292,6 +292,9 @@ func newAudioTrackFromDriver(d driver.Driver, recorder driver.AudioRecorder, con
 		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
 }