Update version to 0.1-rc.6

Refactor docker image and ci

.github/workflows/builder.yml

@@ -1,59 +0,0 @@
-# https://github.com/home-assistant/builder
-name: 'Builder'
-
-on:
-  push:
-    tags: [ 'v*' ]
-  workflow_dispatch:
-
-jobs:
-  hassio:
-    name: Hassio Addon
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout the repository
-        uses: actions/checkout@v3
-
-      - name: Login to DockerHub
-        uses: docker/login-action@v2
-        with:
-          username: ${{ secrets.DOCKERHUB_USERNAME }}
-          password: ${{ secrets.DOCKERHUB_TOKEN }}
-
-      - name: Branch name
-        run: |
-          VERSION="${GITHUB_REF#refs/tags/v}"
-          echo "REPO=alexxit/go2rtc" >> $GITHUB_ENV
-          echo "TAG=${VERSION}" >> $GITHUB_ENV
-          echo "IMAGE=alexxit/go2rtc:${VERSION}" >> $GITHUB_ENV
-
-      - name: Build amd64
-        uses: home-assistant/builder@master
-        with:
-          args: --amd64 --target build/hassio --version $TAG-amd64 --no-latest --docker-hub-check
-
-      - name: Build i386
-        uses: home-assistant/builder@master
-        with:
-          args: --i386 --target build/hassio --version $TAG-i386 --no-latest --docker-hub-check
-
-      - name: Build aarch64
-        uses: home-assistant/builder@master
-        with:
-          args: --aarch64 --target build/hassio --version $TAG-aarch64 --no-latest --docker-hub-check
-
-      - name: Build armv7
-        uses: home-assistant/builder@master
-        with:
-          args: --armv7 --target build/hassio --version $TAG-armv7 --no-latest --docker-hub-check
-
-      - name: Docker manifest
-        run: |
-          # thanks to https://github.com/aler9/rtsp-simple-server/blob/main/Makefile
-          docker manifest create "${IMAGE}" \
-            "${IMAGE}-amd64" "${IMAGE}-i386" "${IMAGE}-aarch64" "${IMAGE}-armv7"
-          docker manifest push "${IMAGE}"
-
-          docker manifest create "${REPO}:latest" \
-            "${IMAGE}-amd64" "${IMAGE}-i386" "${IMAGE}-aarch64" "${IMAGE}-armv7"
-          docker manifest push "${REPO}:latest"

.github/workflows/ci.yml

@@ -0,0 +1,75 @@
+name: ci
+
+on:
+  workflow_dispatch:
+  push:
+    branches:
+      - 'master'
+    tags:
+      - 'v*'
+
+jobs:
+  build-and-push:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v3
+
+      - name: Docker meta
+        id: meta
+        uses: docker/metadata-action@v4
+        with:
+          images: alexxit/go2rtc
+          tags: |
+            type=ref,event=branch
+            type=semver,pattern={{version}},enable=false
+            type=match,pattern=v(.*),group=1
+
+      - name: Docker meta Hardware
+        id: meta-hw
+        uses: docker/metadata-action@v4
+        with:
+          images: alexxit/go2rtc
+          flavor: |
+            suffix=-hardware
+            latest=false
+          tags: |
+            type=ref,event=branch
+            type=semver,pattern={{version}},enable=false
+            type=match,pattern=v(.*),group=1
+
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@v2
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v2
+
+      - name: Login to DockerHub
+        if: github.event_name != 'pull_request'
+        uses: docker/login-action@v2
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
+
+      - name: Build and push
+        uses: docker/build-push-action@v3
+        with:
+          context: .
+          platforms: |
+            linux/amd64
+            linux/386
+            linux/arm/v7
+            linux/arm64/v8
+          push: ${{ github.event_name != 'pull_request' }}
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}
+
+      - name: Build and push Hardware
+        uses: docker/build-push-action@v3
+        with:
+          context: .
+          file: hardware.Dockerfile
+          platforms: linux/amd64
+          push: ${{ github.event_name != 'pull_request' }}
+          tags: ${{ steps.meta-hw.outputs.tags }}
+          labels: ${{ steps.meta-hw.outputs.labels }}

Dockerfile

@@ -0,0 +1,55 @@
+# 0. Prepare images
+ARG PYTHON_VERSION="3.11"
+ARG GO_VERSION="1.19"
+ARG NGROK_VERSION="3"
+
+FROM python:${PYTHON_VERSION}-alpine AS base
+FROM golang:${GO_VERSION}-alpine AS go
+FROM ngrok/ngrok:${NGROK_VERSION}-alpine AS ngrok
+
+
+# 1. Build go2rtc binary
+FROM go AS build
+
+WORKDIR /build
+
+# Cache dependencies
+COPY go.mod go.sum ./
+RUN go mod download
+
+COPY . .
+RUN CGO_ENABLED=0 go build -ldflags "-s -w" -trimpath
+
+
+# 2. Collect all files
+FROM scratch AS rootfs
+
+COPY --from=build /build/go2rtc /usr/local/bin/
+COPY --from=ngrok /bin/ngrok /usr/local/bin/
+COPY ./build/docker/run.sh /
+
+
+# 3. Final image
+FROM base
+
+# Install ffmpeg, bash (for run.sh), tini (for signal handling),
+# and other common tools for the echo source.
+RUN apk add --no-cache tini ffmpeg bash curl jq
+
+# Hardware Acceleration for Intel CPU (+50MB)
+ARG TARGETARCH
+
+RUN if [ "${TARGETARCH}" = "amd64" ]; then apk add --no-cache libva-intel-driver intel-media-driver; fi
+
+# Hardware: AMD and NVidia VAAPI (not sure about this)
+# RUN libva-glx mesa-va-gallium
+# Hardware: AMD and NVidia VDPAU (not sure about this)
+# RUN libva-vdpau-driver mesa-vdpau-gallium (+150MB total)
+
+COPY --from=rootfs / /
+
+RUN chmod a+x /run.sh && mkdir -p /config
+
+ENTRYPOINT ["/sbin/tini", "--"]
+
+CMD ["/run.sh"]

README.md

@@ -6,7 +6,7 @@ Ultimate camera streaming application with support RTSP, WebRTC, HomeKit, FFmpeg
 
 - zero-dependency and zero-config [small app](#go2rtc-binary) for all OS (Windows, macOS, Linux, ARM)
 - zero-delay for many supported protocols (lowest possible streaming latency)
-- streaming from [RTSP](#source-rtsp), [RTMP](#source-rtmp), [MJPEG](#source-ffmpeg), [HLS/HTTP](#source-ffmpeg), [USB Cameras](#source-ffmpeg-device) and [other sources](#module-streams)
+- streaming from [RTSP](#source-rtsp), [RTMP](#source-rtmp), [HTTP](#source-http) (FLV/MJPEG/JPEG), [FFmpeg](#source-ffmpeg), [USB Cameras](#source-ffmpeg-device) and [other sources](#module-streams)
 - streaming to [RTSP](#module-rtsp), [WebRTC](#module-webrtc), [MSE/MP4](#module-mp4) or [MJPEG](#module-mjpeg)
 - first project in the World with support streaming from [HomeKit Cameras](#source-homekit)
 - first project in the World with support H265 for WebRTC in browser ([read more](https://github.com/AlexxIT/Blog/issues/5))
@@ -50,13 +50,14 @@ Download binary for your OS from [latest release](https://github.com/AlexxIT/go2
 
 - `go2rtc_win64.zip` - Windows 64-bit
 - `go2rtc_win32.zip` - Windows 32-bit
+- `go2rtc_win_arm64.zip` - Windows ARM 64-bit
 - `go2rtc_linux_amd64` - Linux 64-bit
 - `go2rtc_linux_i386` - Linux 32-bit
 - `go2rtc_linux_arm64` - Linux ARM 64-bit (ex. Raspberry 64-bit OS)
 - `go2rtc_linux_arm` - Linux ARM 32-bit (ex. Raspberry 32-bit OS)
-- `go2rtc_linux_mipsel` - Linux on MIPS (ex. [Xiaomi Gateway 3](https://github.com/AlexxIT/XiaomiGateway3))
-- `go2rtc_mac_amd64` - Mac with Intel
-- `go2rtc_mac_arm64` - Mac with M1
+- `go2rtc_linux_mipsel` - Linux MIPS (ex. [Xiaomi Gateway 3](https://github.com/AlexxIT/XiaomiGateway3))
+- `go2rtc_mac_amd64.zip` - Mac Intel 64-bit
+- `go2rtc_mac_arm64.zip` - Mac ARM 64-bit
 
 Don't forget to fix the rights `chmod +x go2rtc_xxx_xxx` on Linux and Mac.
 
@@ -329,7 +330,34 @@ More cameras, like [Tuya](https://www.home-assistant.io/integrations/tuya/), [ON
 
 The HTTP API is the main part for interacting with the application. Default address: `http://127.0.0.1:1984/`.
 
-- you can use WebRTC only when HTTP API enabled
+go2rtc has its own JS video player (`video-rtc.js`) with:
+
+- support technologies:
+   - WebRTC over UDP or TCP
+   - MSE or MP4 or MJPEG over WebSocket 
+- automatic selection best technology according on:
+   - codecs inside your stream
+   - current browser capabilities
+   - current network configuration
+- automatic stop stream while browser or page not active
+- automatic stop stream while player not inside page viewport
+- automatic reconnection
+
+Technology selection based on priorities:
+
+1. Video and Audio better than just Video
+2. H265 better than H264
+3. WebRTC better than MSE, than MP4, than MJPEG
+
+go2rtc has simple HTML page (`stream.html`) with support params in URL:
+
+- multiple streams on page `src=camera1&src=camera2...`
+- stream technology autoselection `mode=webrtc,mse,mp4,mjpeg`
+- stream technology comparison `src=camera1&mode=webrtc&mode=mse&mode=mp4`
+- player width setting in pixels `width=320px` or percents `width=50%`
+
+**Module config**
+
 - you can disable HTTP API with `listen: ""` and use, for example, only RTSP client/server protocol
 - you can enable HTTP API only on localhost with `listen: "127.0.0.1:1984"` setting
 - you can change API `base_path` and host go2rtc on your main app webserver suburl
@@ -337,16 +365,20 @@ The HTTP API is the main part for interacting with the application. Default addr
 
 ```yaml
 api:
-  listen: ":1984"    # HTTP API port ("" - disabled)
-  base_path: "/rtc"  # API prefix for serve on suburl (/api => /rtc/api)
-  static_dir: "www"  # folder for static files (custom web interface)
-  origin: "*"        # allow CORS requests (only * supported)
+  listen: ":1984"    # default ":1984", HTTP API port ("" - disabled)
+  base_path: "/rtc"  # default "", API prefix for serve on suburl (/api => /rtc/api)
+  static_dir: "www"  # default "", folder for static files (custom web interface)
+  origin: "*"        # default "", allow CORS requests (only * supported)
 ```
 
 **PS. go2rtc** doesn't provide HTTPS or password protection. Use [Nginx](https://nginx.org/) or [Ngrok](#module-ngrok) or [Home Assistant Add-on](#go2rtc-home-assistant-add-on) for this tasks.
 
 **PS2.** You can access microphone (for 2-way audio) only with HTTPS ([read more](https://stackoverflow.com/questions/52759992/how-to-access-camera-and-microphone-in-chrome-without-https)).
 
+**PS3.** MJPEG over WebSocket plays better than native MJPEG because Chrome [bug](https://bugs.chromium.org/p/chromium/issues/detail?id=527446).
+
+**PS4.** MP4 over WebSocket was created only for Apple iOS because it doesn't support MSE and native MP4.
+
 ### Module: RTSP
 
 You can get any stream as RTSP-stream: `rtsp://192.168.1.123:8554/{stream_name}`
@@ -501,7 +533,8 @@ View almost any Hass camera using `WebRTC` technology, supported codecs `H264`/`
 When the stream starts - the camera `entity_id` will be added to go2rtc "on the fly". You don't need to add cameras manually to [go2rtc config](#configuration). Some cameras (like [Nest](https://www.home-assistant.io/integrations/nest/)) have a dynamic link to the stream, it will be updated each time a stream is started from the Hass interface. 
 
 1. Hass > Settings > Integrations > Add Integration > [RTSPtoWebRTC](https://my.home-assistant.io/redirect/config_flow_start/?domain=rtsp_to_webrtc) > `http://127.0.0.1:1984/`
-2. Use Picture Entity or Picture Glance lovelace card
+2. RTSPtoWebRTC > Configure > STUN server: `stun.l.google.com:19302`
+3. Use Picture Entity or Picture Glance lovelace card
 
 You can add camera `entity_id` to [go2rtc config](#configuration) if you need transcoding:
 

build/docker/run.sh

@@ -0,0 +1,19 @@
+#!/bin/bash
+
+set -euo pipefail
+
+echo "Starting go2rtc..." >&2
+
+readonly config_path="/config"
+
+if [[ -x "${config_path}/go2rtc" ]]; then
+  readonly binary_path="${config_path}/go2rtc"
+  echo "Using go2rtc binary from '${binary_path}' instead of the embedded one" >&2
+else
+  readonly binary_path="/usr/local/bin/go2rtc"
+fi
+
+# set cwd for go2rtc (for config file, Hass integration, etc)
+cd "${config_path}" || echo "Could not change working directory to '${config_path}'" >&2
+
+exec "${binary_path}"

build/hassio/Dockerfile

@@ -1,41 +0,0 @@
-ARG BUILD_FROM
-
-FROM $BUILD_FROM as build
-
-# 1. Build go2rtc
-RUN apk add --no-cache git go
-
-RUN git clone https://github.com/AlexxIT/go2rtc \
-    && cd go2rtc \
-    && CGO_ENABLED=0 go build -ldflags "-s -w" -trimpath
-
-# 2. Download ngrok
-ARG BUILD_ARCH
-
-# https://github.com/home-assistant/docker-base/blob/master/alpine/Dockerfile
-RUN if [ "${BUILD_ARCH}" = "aarch64" ]; then BUILD_ARCH="arm64"; \
-    elif [ "${BUILD_ARCH}" = "armv7" ]; then BUILD_ARCH="arm"; fi \
-    && cd go2rtc \
-    && curl $(curl -s "https://raw.githubusercontent.com/ngrok/docker-ngrok/main/releases.json" | jq -r ".${BUILD_ARCH}.url") -o ngrok.zip \
-    && unzip ngrok
-
-
-
-# https://devopscube.com/reduce-docker-image-size/
-FROM $BUILD_FROM
-
-# 3. Copy go2rtc and ngrok to release
-COPY --from=build /go2rtc/go2rtc /usr/local/bin
-COPY --from=build /go2rtc/ngrok /usr/local/bin
-
-# 4. Install ffmpeg
-# apk base  OK: 22  MiB in 40  packages
-# ffmpeg    OK: 113 MiB in 110 packages
-# python3   OK: 161 MiB in 114 packages
-RUN apk add --no-cache ffmpeg python3
-
-# 5. Copy run to release
-COPY run.sh /
-RUN chmod a+x /run.sh
-
-CMD [ "/run.sh" ]

build/hassio/config.yaml

@@ -1,6 +0,0 @@
-# https://github.com/home-assistant/builder/blob/master/builder.sh
-name: go2rtc
-description: Ultimate camera streaming application
-url: https://github.com/AlexxIT/go2rtc
-image: alexxit/go2rtc
-arch: [ amd64, aarch64, i386, armv7 ]

build/hassio/run.sh

@@ -1,14 +0,0 @@
-#!/usr/bin/with-contenv bashio
-
-set +e
-
-# set cwd for go2rtc (for config file, Hass integration, etc)
-cd /config
-
-# add the feature to override go2rtc binary from Hass config folder
-export PATH="/config:$PATH"
-
-while true; do
-  go2rtc
-  sleep 5
-done

cmd/api/ws.go

@@ -25,7 +25,7 @@ var wsHandlers = make(map[string]WSHandler)
 func initWS(origin string) {
 	wsUp = &websocket.Upgrader{
 		ReadBufferSize:  1024,
-		WriteBufferSize: 512000,
+		WriteBufferSize: 2028,
 	}
 
 	switch origin {

cmd/app/app.go

@@ -10,7 +10,7 @@ import (
 	"runtime"
 )
 
-var Version = "0.1-rc.5"
+var Version = "0.1-rc.6"
 var UserAgent = "go2rtc/" + Version
 
 func Init() {

cmd/ffmpeg/ffmpeg.go

@@ -1,6 +1,8 @@
 package ffmpeg
 
 import (
+	"bytes"
+	"errors"
 	"github.com/AlexxIT/go2rtc/cmd/app"
 	"github.com/AlexxIT/go2rtc/cmd/exec"
 	"github.com/AlexxIT/go2rtc/cmd/ffmpeg/device"
@@ -17,10 +19,29 @@ func Init() {
 		Mod map[string]string `yaml:"ffmpeg"`
 	}
 
-	// defaults
+	cfg.Mod = defaults // will be overriden from yaml
 
-	cfg.Mod = map[string]string{
+	app.LoadConfig(&cfg)
+
+	if app.GetLogger("exec").GetLevel() >= 0 {
+		defaults["global"] += " -v error"
+	}
+
+	streams.HandleFunc("ffmpeg", func(url string) (streamer.Producer, error) {
+		args := parseArgs(url[7:]) // remove `ffmpeg:`
+		if args == nil {
+			return nil, errors.New("can't generate ffmpeg command")
+		}
+		return exec.Handle("exec:" + args.String())
+	})
+
+	device.Bin = defaults["bin"]
+	device.Init()
+}
+
+var defaults = map[string]string{
 	"bin":    "ffmpeg",
+	"global": "-hide_banner",
 
 	// inputs
 	"file": "-re -stream_loop -1 -i {input}",
@@ -30,16 +51,13 @@ func Init() {
 	// output
 	"output": "-user_agent ffmpeg/go2rtc -rtsp_transport tcp -f rtsp {output}",
 
-		// `-g 30` - group of picture, GOP, keyframe interval
 	// `-preset superfast` - we can't use ultrafast because it doesn't support `-profile main -level 4.1`
 	// `-tune zerolatency` - for minimal latency
-		// `-profile main -level 4.1` - most used streaming profile
-		// `-pix_fmt yuv420p` - if input pix format 4:2:2
-		"h264":       "-c:v libx264 -g:v 30 -preset:v superfast -tune:v zerolatency -profile:v main -level:v 4.1 -pix_fmt:v yuv420p",
-		"h264/ultra": "-c:v libx264 -g:v 30 -preset:v ultrafast -tune:v zerolatency",
-		"h264/high":  "-c:v libx264 -g:v 30 -preset:v superfast -tune:v zerolatency",
-		"h265":       "-c:v libx265 -g:v 30 -preset:v superfast -tune:v zerolatency -profile:v main -level:v 5.1 -pix_fmt:v yuv420p",
+	// `-profile high -level 4.1` - most used streaming profile
+	"h264":  "-c:v libx264 -g 50 -profile:v high -level:v 4.1 -preset:v superfast -tune:v zerolatency",
+	"h265":  "-c:v libx265 -g 50 -profile:v high -level:v 5.1 -preset:v superfast -tune:v zerolatency",
 	"mjpeg": "-c:v mjpeg -force_duplicated_matrix:v 1 -huffman:v 0 -pix_fmt:v yuvj420p",
+
 	"opus":       "-c:a libopus -ar:a 48000 -ac:a 2",
 	"pcmu":       "-c:a pcm_mulaw -ar:a 8000 -ac:a 1",
 	"pcmu/16000": "-c:a pcm_mulaw -ar:a 16000 -ac:a 1",
@@ -49,156 +67,176 @@ func Init() {
 	"pcma/48000": "-c:a pcm_alaw -ar:a 48000 -ac:a 1",
 	"aac":        "-c:a aac", // keep sample rate and channels
 	"aac/16000":  "-c:a aac -ar:a 16000 -ac:a 1",
+
+	// hardware Intel and AMD on Linux
+	// better not to set `-async_depth:v 1` like for QSV, because framedrops
+	// `-bf 0` - disable B-frames is very important
+	"h264/vaapi":  "-c:v h264_vaapi -g 50 -bf 0 -profile:v high -level:v 4.1 -sei:v 0",
+	"h265/vaapi":  "-c:v hevc_vaapi -g 50 -bf 0 -profile:v high -level:v 5.1 -sei:v 0",
+	"mjpeg/vaapi": "-c:v mjpeg_vaapi",
+
+	// hardware Raspberry
+	"h264/v4l2m2m": "-c:v h264_v4l2m2m -g 50 -bf 0",
+	"h265/v4l2m2m": "-c:v hevc_v4l2m2m -g 50 -bf 0",
+
+	// hardware NVidia on Linux and Windows
+	// preset=p2 - faster, tune=ll - low latency
+	"h264/cuda": "-c:v h264_nvenc -g 50 -profile:v high -level:v auto -preset:v p2 -tune:v ll",
+	"h265/cuda": "-c:v hevc_nvenc -g 50 -profile:v high -level:v auto",
+
+	// hardware Intel on Windows
+	"h264/dxva2":  "-c:v h264_qsv -g 50 -bf 0 -profile:v high -level:v 4.1 -async_depth:v 1",
+	"h265/dxva2":  "-c:v hevc_qsv -g 50 -bf 0 -profile:v high -level:v 5.1 -async_depth:v 1",
+	"mjpeg/dxva2": "-c:v mjpeg_qsv -profile:v high -level:v 5.1",
+
+	// hardware macOS
+	"h264/videotoolbox": "-c:v h264_videotoolbox -g 50 -bf 0 -profile:v high -level:v 4.1",
+	"h265/videotoolbox": "-c:v hevc_videotoolbox -g 50 -bf 0 -profile:v high -level:v 5.1",
+}
+
+func parseArgs(s string) *Args {
+	// init FFmpeg arguments
+	args := &Args{
+		bin:    defaults["bin"],
+		global: defaults["global"],
+		output: defaults["output"],
 	}
 
-	app.LoadConfig(&cfg)
-
-	tpl := cfg.Mod
-
-	cmd := "exec:" + tpl["bin"] + " -hide_banner "
-
-	if app.GetLogger("exec").GetLevel() >= 0 {
-		cmd += "-v error "
-	}
-
-	streams.HandleFunc("ffmpeg", func(s string) (streamer.Producer, error) {
-		s = s[7:] // remove `ffmpeg:`
-
 	var query url.Values
-		var queryVideo, queryAudio bool
-
 	if i := strings.IndexByte(s, '#'); i > 0 {
 		query = parseQuery(s[i+1:])
-			queryVideo = query["video"] != nil
-			queryAudio = query["audio"] != nil
+		args.video = len(query["video"])
+		args.audio = len(query["audio"])
 		s = s[:i]
-		} else {
-			// by default query both video and audio
-			queryVideo = true
-			queryAudio = true
 	}
 
-		var input string
+	// Parse input:
+	//   1. Input as xxxx:// link (http or rtsp or any other)
+	//   2. Input as stream name
+	//   3. Input as FFmpeg device (local USB camera)
 	if i := strings.Index(s, "://"); i > 0 {
 		switch s[:i] {
 		case "http", "https", "rtmp":
-				input = strings.Replace(tpl["http"], "{input}", s, 1)
+			args.input = strings.Replace(defaults["http"], "{input}", s, 1)
 		case "rtsp", "rtsps":
 			// https://ffmpeg.org/ffmpeg-protocols.html#rtsp
 			// skip unnecessary input tracks
 			switch {
-				case queryVideo && queryAudio:
-					input = "-allowed_media_types video+audio "
-				case queryVideo:
-					input = "-allowed_media_types video "
-				case queryAudio:
-					input = "-allowed_media_types audio "
+			case (args.video > 0 && args.audio > 0) || (args.video == 0 && args.audio == 0):
+				args.input = "-allowed_media_types video+audio "
+			case args.video > 0:
+				args.input = "-allowed_media_types video "
+			case args.audio > 0:
+				args.input = "-allowed_media_types audio "
 			}
 
-				input += strings.Replace(tpl["rtsp"], "{input}", s, 1)
+			args.input += strings.Replace(defaults["rtsp"], "{input}", s, 1)
 		default:
-				input = "-i " + s
+			args.input = "-i " + s
 		}
 	} else if streams.Get(s) != nil {
 		s = "rtsp://localhost:" + rtsp.Port + "/" + s
 		switch {
-			case queryVideo && !queryAudio:
+		case args.video > 0 && args.audio == 0:
 			s += "?video"
-			case queryAudio && !queryVideo:
+		case args.audio > 0 && args.video == 0:
 			s += "?audio"
 		}
-			input = strings.Replace(tpl["rtsp"], "{input}", s, 1)
+		args.input = strings.Replace(defaults["rtsp"], "{input}", s, 1)
 	} else if strings.HasPrefix(s, "device?") {
 		var err error
-			input, err = device.GetInput(s)
+		args.input, err = device.GetInput(s)
 		if err != nil {
-				return nil, err
+			return nil
 		}
 	} else {
-			input = strings.Replace(tpl["file"], "{input}", s, 1)
+		args.input = strings.Replace(defaults["file"], "{input}", s, 1)
 	}
 
-		if _, ok := query["async"]; ok {
-			input = "-use_wallclock_as_timestamps 1 -async 1 " + input
+	if query["async"] != nil {
+		args.input = "-use_wallclock_as_timestamps 1 -async 1 " + args.input
 	}
 
-		s = cmd + input
-
+	// Parse query params:
+	//   1. `width`/`height` params
+	//   2. `rotate` param
+	//   3. `video` params (support multiple)
+	//   4. `audio` params (support multiple)
+	//   5. `hardware` param
 	if query != nil {
+		// 1. Process raw params for FFmpeg
 		for _, raw := range query["raw"] {
-				s += " " + raw
+			args.AddCodec(raw)
 		}
 
-			for _, rotate := range query["rotate"] {
-				switch rotate {
+		// 2. Process video filters (resize and rotation)
+		if query["width"] != nil || query["height"] != nil {
+			filter := "scale="
+			if query["width"] != nil {
+				filter += query["width"][0]
+			} else {
+				filter += "-1"
+			}
+			filter += ":"
+			if query["height"] != nil {
+				filter += query["height"][0]
+			} else {
+				filter += "-1"
+			}
+			args.AddFilter(filter)
+		}
+
+		if query["rotate"] != nil {
+			var filter string
+			switch query["rotate"][0] {
 			case "90":
-					s += " -vf transpose=1" // 90 degrees clockwise
+				filter = "transpose=1" // 90 degrees clockwise
 			case "180":
-					s += " -vf transpose=1,transpose=1"
+				filter = "transpose=1,transpose=1"
 			case "-90", "270":
-					s += " -vf transpose=2" // 90 degrees counterclockwise
+				filter = "transpose=2" // 90 degrees counterclockwise
+			}
+			if filter != "" {
+				args.AddFilter(filter)
 			}
-				break
 		}
 
-			switch len(query["video"]) {
-			case 0:
-				s += " -vn"
-			case 1:
-				if len(query["audio"]) > 1 {
-					s += " -map 0:v:0?"
-				}
+		// 3. Process video codecs
+		if args.video > 0 {
 			for _, video := range query["video"] {
-					if video == "copy" {
-						s += " -c:v copy"
+				if video != "copy" {
+					args.AddCodec(defaults[video])
 				} else {
-						s += " " + tpl[video]
+					args.AddCodec("-c:v copy")
 				}
 			}
-			default:
-				for i, video := range query["video"] {
-					if video == "copy" {
-						s += " -map 0:v:0? -c:v:" + strconv.Itoa(i) + " copy"
 		} else {
-						s += " -map 0:v:0? " + strings.ReplaceAll(tpl[video], ":v ", ":v:"+strconv.Itoa(i)+" ")
-					}
-				}
+			args.AddCodec("-vn")
 		}
 
-			switch len(query["audio"]) {
-			case 0:
-				s += " -an"
-			case 1:
-				if len(query["video"]) > 1 {
-					s += " -map 0:a:0?"
-				}
+		// 4. Process audio codecs
+		if args.audio > 0 {
 			for _, audio := range query["audio"] {
-					if audio == "copy" {
-						s += " -c:a copy"
+				if audio != "copy" {
+					args.AddCodec(defaults[audio])
 				} else {
-						s += " " + tpl[audio]
-					}
-				}
-			default:
-				for i, audio := range query["audio"] {
-					if audio == "copy" {
-						s += " -map 0:a:0? -c:a:" + strconv.Itoa(i) + " copy"
-					} else {
-						s += " -map 0:a:0? " + strings.ReplaceAll(tpl[audio], ":a ", ":a:"+strconv.Itoa(i)+" ")
-					}
+					args.AddCodec("-c:a copy")
 				}
 			}
 		} else {
-			s += " -c copy"
+			args.AddCodec("-an")
 		}
 
-		s += " " + tpl["output"]
+		if query["hardware"] != nil {
+			MakeHardware(args, query["hardware"][0])
+		}
+	}
 
-		return exec.Handle(s)
-	})
+	if args.codecs == nil {
+		args.AddCodec("-c copy")
+	}
 
-	device.Bin = cfg.Mod["bin"]
-	device.Init()
+	return args
 }
 
 func parseQuery(s string) map[string][]string {
@@ -213,3 +251,76 @@ func parseQuery(s string) map[string][]string {
 	}
 	return query
 }
+
+type Args struct {
+	bin     string   // ffmpeg
+	global  string   // -hide_banner -v error
+	input   string   // -re -stream_loop -1 -i /media/bunny.mp4
+	codecs  []string // -c:v libx264 -g:v 30 -preset:v ultrafast -tune:v zerolatency
+	filters []string // scale=1920:1080
+	output  string   // -f rtsp {output}
+
+	video, audio int // count of video and audio params
+}
+
+func (a *Args) AddCodec(codec string) {
+	a.codecs = append(a.codecs, codec)
+}
+
+func (a *Args) AddFilter(filter string) {
+	a.filters = append(a.filters, filter)
+}
+
+func (a *Args) InsertFilter(filter string) {
+	a.filters = append([]string{filter}, a.filters...)
+}
+
+func (a *Args) String() string {
+	b := bytes.NewBuffer(make([]byte, 0, 512))
+
+	b.WriteString(a.bin)
+
+	if a.global != "" {
+		b.WriteByte(' ')
+		b.WriteString(a.global)
+	}
+
+	b.WriteByte(' ')
+	b.WriteString(a.input)
+
+	multimode := a.video > 1 || a.audio > 1
+	var iv, ia int
+
+	for _, codec := range a.codecs {
+		// support multiple video and/or audio codecs
+		if multimode && len(codec) >= 5 {
+			switch codec[:5] {
+			case "-c:v ":
+				codec = "-map 0:v:0? " + strings.ReplaceAll(codec, ":v ", ":v:"+strconv.Itoa(iv)+" ")
+				iv++
+			case "-c:a ":
+				codec = "-map 0:a:0? " + strings.ReplaceAll(codec, ":a ", ":a:"+strconv.Itoa(ia)+" ")
+				ia++
+			}
+		}
+
+		b.WriteByte(' ')
+		b.WriteString(codec)
+	}
+
+	if a.filters != nil {
+		for i, filter := range a.filters {
+			if i == 0 {
+				b.WriteString(" -vf ")
+			} else {
+				b.WriteByte(',')
+			}
+			b.WriteString(filter)
+		}
+	}
+
+	b.WriteByte(' ')
+	b.WriteString(a.output)
+
+	return b.String()
+}

cmd/ffmpeg/hardware.go

@@ -0,0 +1,112 @@
+package ffmpeg
+
+import (
+	"github.com/rs/zerolog/log"
+	"os/exec"
+	"strings"
+)
+
+const (
+	EngineSoftware     = "software"
+	EngineVAAPI        = "vaapi"        // Intel iGPU and AMD GPU
+	EngineV4L2M2M      = "v4l2m2m"      // Raspberry Pi 3 and 4
+	EngineCUDA         = "cuda"         // NVidia on Windows and Linux
+	EngineDXVA2        = "dxva2"        // Intel on Windows
+	EngineVideoToolbox = "videotoolbox" // macOS
+)
+
+var cache = map[string]string{}
+
+// MakeHardware converts software FFmpeg args to hardware args
+// empty engine for autoselect
+func MakeHardware(args *Args, engine string) {
+	for i, codec := range args.codecs {
+		if len(codec) < 12 {
+			continue // skip short line (-c:v libx264...)
+		}
+
+		// get current codec name
+		name := cut(codec, ' ', 1)
+		switch name {
+		case "libx264":
+			name = "h264"
+		case "libx265":
+			name = "h265"
+		case "mjpeg":
+		default:
+			continue // skip unsupported codec
+		}
+
+		// temporary disable probe for H265 and MJPEG
+		if engine == "" && name == "h264" {
+			if engine = cache[name]; engine == "" {
+				engine = ProbeHardware(name)
+				cache[name] = engine
+			}
+		}
+
+		switch engine {
+		case EngineVAAPI:
+			args.input = "-hwaccel vaapi -hwaccel_output_format vaapi " + args.input
+			args.codecs[i] = defaults[name+"/"+engine]
+
+			for i, filter := range args.filters {
+				if strings.HasPrefix(filter, "scale=") {
+					args.filters[i] = "scale_vaapi=" + filter[6:]
+				}
+			}
+
+			// fix if input doesn't support hwaccel, do nothing when support
+			args.InsertFilter("format=vaapi|nv12,hwupload")
+
+		case EngineCUDA:
+			args.input = "-hwaccel cuda -hwaccel_output_format cuda -extra_hw_frames 2 " + args.input
+			args.codecs[i] = defaults[name+"/"+engine]
+
+			for i, filter := range args.filters {
+				if strings.HasPrefix(filter, "scale=") {
+					args.filters[i] = "scale_cuda=" + filter[6:]
+				}
+			}
+
+		case EngineDXVA2:
+			args.input = "-hwaccel dxva2 -hwaccel_output_format dxva2_vld " + args.input
+			args.codecs[i] = defaults[name+"/"+engine]
+
+			for i, filter := range args.filters {
+				if strings.HasPrefix(filter, "scale=") {
+					args.filters[i] = "scale_qsv=" + filter[6:]
+				}
+			}
+
+			args.InsertFilter("hwmap=derive_device=qsv,format=qsv")
+
+		case EngineVideoToolbox:
+			args.input = "-hwaccel videotoolbox -hwaccel_output_format videotoolbox_vld " + args.input
+			args.codecs[i] = defaults[name+"/"+engine]
+
+		case EngineV4L2M2M:
+			args.codecs[i] = defaults[name+"/"+engine]
+		}
+	}
+}
+
+func run(arg ...string) bool {
+	err := exec.Command(defaults["bin"], arg...).Run()
+	log.Printf("%v %v", arg, err)
+	return err == nil
+}
+
+func cut(s string, sep byte, pos int) string {
+	for n := 0; n < pos; n++ {
+		if i := strings.IndexByte(s, sep); i > 0 {
+			s = s[i+1:]
+		} else {
+			return ""
+		}
+	}
+	if i := strings.IndexByte(s, sep); i > 0 {
+		return s[:i]
+	}
+	return s
+}

cmd/ffmpeg/hardware_darwin.go

@@ -0,0 +1,21 @@
+package ffmpeg
+
+func ProbeHardware(name string) string {
+	switch name {
+	case "h264":
+		if run(
+			"-f", "lavfi", "-i", "testsrc2", "-t", "1",
+			"-c", "h264_videotoolbox", "-f", "null", "-") {
+			return EngineVideoToolbox
+		}
+
+	case "h265":
+		if run(
+			"-f", "lavfi", "-i", "testsrc2", "-t", "1",
+			"-c", "hevc_videotoolbox", "-f", "null", "-") {
+			return EngineVideoToolbox
+		}
+	}
+
+	return EngineSoftware
+}

cmd/ffmpeg/hardware_linux.go

@@ -0,0 +1,67 @@
+package ffmpeg
+
+import (
+	"runtime"
+)
+
+func ProbeHardware(name string) string {
+	if runtime.GOARCH == "arm64" || runtime.GOARCH == "arm" {
+		switch name {
+		case "h264":
+			if run(
+				"-f", "lavfi", "-i", "testsrc2", "-t", "1",
+				"-c", "h264_v4l2m2m", "-f", "null", "-") {
+				return EngineV4L2M2M
+			}
+
+		case "h265":
+			if run(
+				"-f", "lavfi", "-i", "testsrc2", "-t", "1",
+				"-c", "hevc_v4l2m2m", "-f", "null", "-") {
+				return EngineV4L2M2M
+			}
+		}
+
+		return EngineSoftware
+	}
+
+	switch name {
+	case "h264":
+		if run("-init_hw_device", "cuda",
+			"-f", "lavfi", "-i", "testsrc2", "-t", "1",
+			"-c", "h264_nvenc", "-f", "null", "-") {
+			return EngineCUDA
+		}
+
+		if run("-init_hw_device", "vaapi",
+			"-f", "lavfi", "-i", "testsrc2", "-t", "1",
+			"-vf", "format=nv12,hwupload",
+			"-c", "h264_vaapi", "-f", "null", "-") {
+			return EngineVAAPI
+		}
+
+	case "h265":
+		if run("-init_hw_device", "cuda",
+			"-f", "lavfi", "-i", "testsrc2", "-t", "1",
+			"-c", "hevc_nvenc", "-f", "null", "-") {
+			return EngineCUDA
+		}
+
+		if run("-init_hw_device", "vaapi",
+			"-f", "lavfi", "-i", "testsrc2", "-t", "1",
+			"-vf", "format=nv12,hwupload",
+			"-c", "hevc_vaapi", "-f", "null", "-") {
+			return EngineVAAPI
+		}
+
+	case "mjpeg":
+		if run("-init_hw_device", "vaapi",
+			"-f", "lavfi", "-i", "testsrc2", "-t", "1",
+			"-vf", "format=nv12,hwupload",
+			"-c", "mjpeg_vaapi", "-f", "null", "-") {
+			return EngineVAAPI
+		}
+	}
+
+	return EngineSoftware
+}

cmd/ffmpeg/hardware_windows.go

@@ -0,0 +1,40 @@
+package ffmpeg
+
+func ProbeHardware(name string) string {
+	switch name {
+	case "h264":
+		if run("-init_hw_device", "cuda",
+			"-f", "lavfi", "-i", "testsrc2", "-t", "1",
+			"-c", "h264_nvenc", "-f", "null", "-") {
+			return EngineCUDA
+		}
+
+		if run("-init_hw_device", "dxva2",
+			"-f", "lavfi", "-i", "testsrc2", "-t", "1",
+			"-c", "h264_qsv", "-f", "null", "-") {
+			return EngineDXVA2
+		}
+
+	case "h265":
+		if run("-init_hw_device", "cuda",
+			"-f", "lavfi", "-i", "testsrc2", "-t", "1",
+			"-c", "hevc_nvenc", "-f", "null", "-") {
+			return EngineCUDA
+		}
+
+		if run("-init_hw_device", "dxva2",
+			"-f", "lavfi", "-i", "testsrc2", "-t", "1",
+			"-c", "hevc_qsv", "-f", "null", "-") {
+			return EngineDXVA2
+		}
+
+	case "mjpeg":
+		if run("-init_hw_device", "dxva2",
+			"-f", "lavfi", "-i", "testsrc2", "-t", "1",
+			"-c", "mjpeg_qsv", "-f", "null", "-") {
+			return EngineDXVA2
+		}
+	}
+
+	return EngineSoftware
+}

cmd/mjpeg/mjpeg.go

@@ -121,6 +121,8 @@ func handlerWS(tr *api.Transport, _ *api.Message) error {
 		return err
 	}
 
+	tr.Write(&api.Message{Type: "mjpeg"})
+
 	tr.OnClose(func() {
 		stream.RemoveConsumer(cons)
 	})

cmd/mp4/ws.go

@@ -9,7 +9,7 @@ import (
 	"strings"
 )
 
-const packetSize = 8192
+const packetSize = 1400
 
 func handlerWSMSE(tr *api.Transport, msg *api.Message) error {
 	src := tr.Request.URL.Query().Get("src")
@@ -86,6 +86,8 @@ func handlerWSMP4(tr *api.Transport, msg *api.Message) error {
 		return err
 	}
 
+	tr.Write(&api.Message{Type: "mp4", Value: cons.MimeType})
+
 	tr.OnClose(func() {
 		stream.RemoveConsumer(cons)
 	})

cmd/streams/producer.go

@@ -24,6 +24,7 @@ type Producer struct {
 	template string
 
 	element streamer.Producer
+	lastErr error
 	tracks  []*streamer.Track
 
 	state   state
@@ -45,10 +46,9 @@ func (p *Producer) GetMedias() []*streamer.Media {
 	if p.state == stateNone {
 		log.Debug().Msgf("[streams] probe producer url=%s", p.url)
 
-		var err error
-		p.element, err = GetProducer(p.url)
-		if err != nil || p.element == nil {
-			log.Error().Err(err).Caller().Send()
+		p.element, p.lastErr = GetProducer(p.url)
+		if p.lastErr != nil || p.element == nil {
+			log.Error().Err(p.lastErr).Caller().Send()
 			return nil
 		}
 

cmd/streams/stream.go

@@ -99,11 +99,18 @@ func (s *Stream) AddConsumer(cons streamer.Consumer) (err error) {
 
 	if len(producers) == 0 {
 		s.stopProducers()
+
 		if len(codecs) > 0 {
 			return errors.New("codecs not match: " + codecs)
-		} else {
-			return fmt.Errorf("sources unavailable: %d", len(s.producers))
 		}
+
+		for i, producer := range s.producers {
+			if producer.lastErr != nil {
+				return fmt.Errorf("source %d error: %w", i, producer.lastErr)
+			}
+		}
+
+		return fmt.Errorf("sources unavailable: %d", len(s.producers))
 	}
 
 	s.mu.Lock()

cmd/webrtc/webrtc.go

@@ -68,7 +68,7 @@ var NewPConn func() (*pion.PeerConnection, error)
 
 func asyncHandler(tr *api.Transport, msg *api.Message) error {
 	src := tr.Request.URL.Query().Get("src")
-	stream := streams.Get(src)
+	stream := streams.GetOrNew(src)
 	if stream == nil {
 		return errors.New(api.StreamNotFound)
 	}

hardware.Dockerfile

@@ -0,0 +1,52 @@
+# 0. Prepare images
+# only debian 12 (bookworm) has latest ffmpeg
+ARG DEBIAN_VERSION="bookworm-slim"
+ARG GO_VERSION="1.19-buster"
+ARG NGROK_VERSION="3"
+
+FROM debian:${DEBIAN_VERSION} AS base
+FROM golang:${GO_VERSION} AS go
+FROM ngrok/ngrok:${NGROK_VERSION} AS ngrok
+
+
+# 1. Build go2rtc binary
+FROM go AS build
+
+WORKDIR /build
+
+# Cache dependencies
+COPY go.mod go.sum ./
+RUN go mod download
+
+COPY . .
+RUN CGO_ENABLED=0 go build -ldflags "-s -w" -trimpath
+
+
+# 2. Collect all files
+FROM scratch AS rootfs
+
+COPY --from=build /build/go2rtc /usr/local/bin/
+COPY --from=ngrok /bin/ngrok /usr/local/bin/
+COPY ./build/docker/run.sh /
+
+
+# 3. Final image
+FROM base
+
+# Install ffmpeg, bash (for run.sh), tini (for signal handling),
+# and other common tools for the echo source.
+# non-free for Intel QSV support (not used by go2rtc, just for tests)
+RUN echo 'deb http://deb.debian.org/debian bookworm non-free' > /etc/apt/sources.list.d/debian-non-free.list && \
+    apt-get -y update && apt-get -y install tini ffmpeg python3 curl jq intel-media-va-driver-non-free
+
+COPY --from=rootfs / /
+
+RUN chmod a+x /run.sh && mkdir -p /config
+
+ENTRYPOINT ["/usr/bin/tini", "--"]
+
+# https://github.com/NVIDIA/nvidia-docker/wiki/Installation-(Native-GPU-Support)
+ENV NVIDIA_VISIBLE_DEVICES all
+ENV NVIDIA_DRIVER_CAPABILITIES compute,video,utility
+
+CMD ["/run.sh"]

pkg/h264/rtp.go

@@ -36,6 +36,7 @@ func RTPDepay(track *streamer.Track) streamer.WrapperFunc {
 			}
 
 			if len(buf) == 0 {
+				for {
 					// Amcrest IP4M-1051: 9, 7, 8, 6, 28...
 					// Amcrest IP4M-1051: 9, 6, 1
 					switch NALUType(payload) {
@@ -52,10 +53,9 @@ func RTPDepay(track *streamer.Track) streamer.WrapperFunc {
 						}
 
 						payload = payload[i:]
-
-					if NALUType(payload) == NALUTypeIFrame {
-						buf = append(buf, ps...)
+						continue
 					}
+					break
 				}
 			}
 

pkg/mjpeg/rfc2435.go

@@ -138,9 +138,9 @@ var chm_ac_symbols = []byte{
 	0xf9, 0xfa,
 }
 
-func MakeHeaders(t byte, w, h uint16, lqt, cqt []byte) []byte {
+func MakeHeaders(p []byte, t byte, w, h uint16, lqt, cqt []byte) []byte {
 	// Appendix A from https://www.rfc-editor.org/rfc/rfc2435
-	p := []byte{0xFF, 0xD8}
+	p = append(p, 0xFF, 0xD8)
 
 	p = MakeQuantHeader(p, lqt, 0)
 	p = MakeQuantHeader(p, cqt, 1)

pkg/mjpeg/rtp.go

@@ -6,7 +6,7 @@ import (
 )
 
 func RTPDepay(track *streamer.Track) streamer.WrapperFunc {
-	var header, payload []byte
+	buf := make([]byte, 0, 512*1024) // 512K
 
 	return func(push streamer.WriterFunc) streamer.WriterFunc {
 		return func(packet *rtp.Packet) error {
@@ -25,7 +25,7 @@ func RTPDepay(track *streamer.Track) streamer.WrapperFunc {
 				b = b[8:]
 			}
 
-			if header == nil {
+			if len(buf) == 0 {
 				var lqt, cqt []byte
 
 				// 3.1.8.  Quantization Table header
@@ -49,26 +49,26 @@ func RTPDepay(track *streamer.Track) streamer.WrapperFunc {
 				}
 
 				//fmt.Printf("t: %d, q: %d, w: %d, h: %d\n", t, q, w, h)
-				header = MakeHeaders(t, w, h, lqt, cqt)
+				buf = MakeHeaders(buf, t, w, h, lqt, cqt)
 			}
 
 			// 3.1.9.  JPEG Payload
-			payload = append(payload, b...)
+			buf = append(buf, b...)
 
 			if !packet.Marker {
 				return nil
 			}
 
-			b = append(header, payload...)
-			if end := b[len(b)-2:]; end[0] != 0xFF && end[1] != 0xD9 {
-				b = append(b, 0xFF, 0xD9)
+			if end := buf[len(buf)-2:]; end[0] != 0xFF && end[1] != 0xD9 {
+				buf = append(buf, 0xFF, 0xD9)
 			}
 
-			header = nil
-			payload = nil
+			clone := *packet
+			clone.Payload = buf
 
-			packet.Payload = b
-			return push(packet)
+			buf = buf[:0] // clear buffer
+
+			return push(&clone)
 		}
 	}
 }

scripts/build.cmd

@@ -3,17 +3,17 @@
 @SET GOOS=windows
 @SET GOARCH=amd64
 @SET FILENAME=go2rtc_win64.zip
-go build -ldflags "-s -w" -trimpath && 7z a -sdel %FILENAME% go2rtc.exe
+go build -ldflags "-s -w" -trimpath && 7z a -mx9 -sdel %FILENAME% go2rtc.exe
 
 @SET GOOS=windows
 @SET GOARCH=386
 @SET FILENAME=go2rtc_win32.zip
-go build -ldflags "-s -w" -trimpath && 7z a -sdel %FILENAME% go2rtc.exe
+go build -ldflags "-s -w" -trimpath && 7z a -mx9 -sdel %FILENAME% go2rtc.exe
 
 @SET GOOS=windows
 @SET GOARCH=arm64
 @SET FILENAME=go2rtc_win_arm64.zip
-go build -ldflags "-s -w" -trimpath && 7z a -sdel %FILENAME% go2rtc.exe
+go build -ldflags "-s -w" -trimpath && 7z a -mx9 -sdel %FILENAME% go2rtc.exe
 
 @SET GOOS=linux
 @SET GOARCH=amd64
@@ -44,9 +44,9 @@ go build -ldflags "-s -w" -trimpath -o %FILENAME% && upx %FILENAME%
 @SET GOOS=darwin
 @SET GOARCH=amd64
 @SET FILENAME=go2rtc_mac_amd64.zip
-go build -ldflags "-s -w" -trimpath && 7z a -sdel %FILENAME% go2rtc
+go build -ldflags "-s -w" -trimpath && 7z a -mx9 -sdel %FILENAME% go2rtc
 
 @SET GOOS=darwin
 @SET GOARCH=arm64
 @SET FILENAME=go2rtc_mac_arm64.zip
-go build -ldflags "-s -w" -trimpath && 7z a -sdel %FILENAME% go2rtc
+go build -ldflags "-s -w" -trimpath && 7z a -mx9 -sdel %FILENAME% go2rtc

www/stream.html

@@ -3,13 +3,13 @@
 <head>
     <meta name="viewport" content="width=device-width, initial-scale=1">
     <title>go2rtc - Stream</title>
-    <script src="video-rtc.js"></script>
     <style>
         body {
             background: black;
             margin: 0;
             padding: 0;
             display: flex;
+            font-family: Arial, Helvetica, sans-serif;
         }
 
         html, body {
@@ -25,7 +25,8 @@
     </style>
 </head>
 <body>
-<script>
+<script type="module" src="./video-stream.js"></script>
+<script type="module">
     const params = new URLSearchParams(location.search);
 
     // support multiple streams and multiple modes
@@ -44,16 +45,16 @@
         document.body.className = "flex";
     }
 
-    const background = params.get("background") === "true";
+    const background = params.get("background") !== "false";
     const width = "1 0 " + (params.get("width") || "320px");
 
     for (let i = 0; i < streams.length; i++) {
-        /** @type {VideoRTC} */
-        const video = document.createElement("video-rtc");
+        /** @type {VideoStream} */
+        const video = document.createElement("video-stream");
         video.background = background;
         video.mode = modes[i] || video.mode;
         video.style.flex = width;
-        video.src = new URL("api/ws?src=" + streams[i], location.href);
+        video.src = new URL("api/ws?src=" + encodeURIComponent(streams[i]), location.href);
         document.body.appendChild(video);
     }
 </script>

www/video-rtc.js

@@ -1,5 +1,5 @@
 /**
- * Video player for MSE and WebRTC connections.
+ * Video player for go2rtc streaming application.
  *
  * All modern web technologies are supported in almost any browser except Apple Safari.
  *
@@ -12,7 +12,7 @@
  * - Customized built-in elements (extends HTMLVideoElement) because all Safari
  * - Public class fields because old Safari (before 14.0)
  */
-class VideoRTC extends HTMLElement {
+export class VideoRTC extends HTMLElement {
     constructor() {
         super();
 
@@ -31,32 +31,50 @@ class VideoRTC extends HTMLElement {
         ];
 
         /**
-         * Supported modes (webrtc, mse, mp4, mjpeg).
+         * [config] Supported modes (webrtc, mse, mp4, mjpeg).
          * @type {string}
          */
         this.mode = "webrtc,mse,mp4,mjpeg";
 
         /**
-         * Run stream when not displayed on the screen. Default `false`.
+         * [config] Run stream when not displayed on the screen. Default `false`.
          * @type {boolean}
          */
         this.background = false;
 
         /**
-         * Run stream only when player in the viewport. Stop when user scroll out player.
+         * [config] Run stream only when player in the viewport. Stop when user scroll out player.
          * Value is percentage of visibility from `0` (not visible) to `1` (full visible).
          * Default `0` - disable;
          * @type {number}
          */
-        this.intersectionThreshold = 0;
+        this.visibilityThreshold = 0;
 
         /**
-         * Run stream only when browser page on the screen. Stop when user change browser
+         * [config] Run stream only when browser page on the screen. Stop when user change browser
          * tab or minimise browser windows.
          * @type {boolean}
          */
         this.visibilityCheck = true;
 
+        /**
+         * [config] WebRTC configuration
+         * @type {RTCConfiguration}
+         */
+        this.pcConfig = {iceServers: [{urls: "stun:stun.l.google.com:19302"}]};
+
+        /**
+         * [info] WebSocket connection state. Values: CONNECTING, OPEN, CLOSED
+         * @type {number}
+         */
+        this.wsState = WebSocket.CLOSED;
+
+        /**
+         * [info] WebRTC connection state.
+         * @type {number}
+         */
+        this.pcState = WebSocket.CLOSED;
+
         /**
          * @type {HTMLVideoElement}
          */
@@ -68,16 +86,9 @@ class VideoRTC extends HTMLElement {
         this.ws = null;
 
         /**
-         * Internal WebSocket connection state. Values: CONNECTING, OPEN, CLOSED
-         * @type {number}
-         */
-        this.wsState = WebSocket.CLOSED;
-
-        /**
-         * Internal WebSocket URL.
          * @type {string|URL}
          */
-        this.url = "";
+        this.wsURL = "";
 
         /**
          * @type {RTCPeerConnection}
@@ -87,37 +98,38 @@ class VideoRTC extends HTMLElement {
         /**
          * @type {number}
          */
-        this.pcState = WebSocket.CLOSED;
-
-        this.pcConfig = {iceServers: [{urls: "stun:stun.l.google.com:19302"}]};
+        this.connectTS = 0;
 
         /**
-         * Internal disconnect TimeoutID.
+         * @type {string}
+         */
+        this.mseCodecs = "";
+
+        /**
+         * [internal] Disconnect TimeoutID.
          * @type {number}
          */
-        this.disconnectTimeout = 0;
+        this.disconnectTID = 0;
 
         /**
-         * Internal reconnect TimeoutID.
+         * [internal] Reconnect TimeoutID.
          * @type {number}
          */
-        this.reconnectTimeout = 0;
+        this.reconnectTID = 0;
 
         /**
-         * Handler for receiving Binary from WebSocket
+         * [internal] Handler for receiving Binary from WebSocket.
          * @type {Function}
          */
         this.ondata = null;
 
         /**
-         * Handlers list for receiving JSON from WebSocket
+         * [internal] Handlers list for receiving JSON from WebSocket
          * @type {Object.<string,Function>}}
          */
         this.onmessage = null;
     }
 
-    /** public properties **/
-
     /**
      * Set video source (WebSocket URL). Support relative path.
      * @param {string|URL} value
@@ -130,9 +142,9 @@ class VideoRTC extends HTMLElement {
             value = "ws" + location.origin.substring(4) + value;
         }
 
-        this.url = value;
+        this.wsURL = value;
 
-        if (this.isConnected) this.connectedCallback();
+        this.onconnect();
     }
 
     /**
@@ -156,10 +168,6 @@ class VideoRTC extends HTMLElement {
         if (this.ws) this.ws.send(JSON.stringify(value));
     }
 
-    get closed() {
-        return this.wsState === WebSocket.CLOSED && this.pcState === WebSocket.CLOSED;
-    }
-
     codecs(type) {
         const test = type === "mse"
             ? codec => MediaSource.isTypeSupported(`video/mp4; codecs="${codec}"`)
@@ -172,11 +180,9 @@ class VideoRTC extends HTMLElement {
      * document-connected element.
      */
     connectedCallback() {
-        console.debug("VideoRTC.connectedCallback", this.wsState, this.pcState);
-
-        if (this.disconnectTimeout) {
-            clearTimeout(this.disconnectTimeout);
-            this.disconnectTimeout = 0;
+        if (this.disconnectTID) {
+            clearTimeout(this.disconnectTID);
+            this.disconnectTID = 0;
         }
 
         // because video autopause on disconnected from DOM
@@ -186,15 +192,11 @@ class VideoRTC extends HTMLElement {
                 this.video.currentTime = seek.end(seek.length - 1);
                 this.play();
             }
+        } else {
+            this.oninit();
         }
 
-        if (!this.url || !this.closed) return;
-
-        // CLOSED => CONNECTING
-        this.wsState = WebSocket.CONNECTING;
-
-        this.internalInit();
-        this.internalWS();
+        this.onconnect();
     }
 
     /**
@@ -202,35 +204,25 @@ class VideoRTC extends HTMLElement {
      * document's DOM.
      */
     disconnectedCallback() {
-        console.debug("VideoRTC.disconnectedCallback", this.wsState, this.pcState);
+        if (this.background || this.disconnectTID) return;
+        if (this.wsState === WebSocket.CLOSED && this.pcState === WebSocket.CLOSED) return;
 
-        if (this.background || this.disconnectTimeout || this.closed) return;
-
-        this.disconnectTimeout = setTimeout(() => {
-            if (this.reconnectTimeout) {
-                clearTimeout(this.reconnectTimeout);
-                this.reconnectTimeout = 0;
+        this.disconnectTID = setTimeout(() => {
+            if (this.reconnectTID) {
+                clearTimeout(this.reconnectTID);
+                this.reconnectTID = 0;
             }
 
-            this.disconnectTimeout = 0;
+            this.disconnectTID = 0;
 
-            this.wsState = WebSocket.CLOSED;
-            if (this.ws) {
-                this.ws.close();
-                this.ws = null;
-            }
-
-            this.pcState = WebSocket.CLOSED;
-            if (this.pc) {
-                this.pc.close();
-                this.pc = null;
-            }
+            this.ondisconnect();
         }, this.DISCONNECT_TIMEOUT);
     }
 
-    internalInit() {
-        if (this.childElementCount) return;
-
+    /**
+     * Creates child DOM elements. Called automatically once on `connectedCallback`.
+     */
+    oninit() {
         this.video = document.createElement("video");
         this.video.controls = true;
         this.video.playsInline = true;
@@ -258,7 +250,7 @@ class VideoRTC extends HTMLElement {
             })
         }
 
-        if ("IntersectionObserver" in window && this.intersectionThreshold) {
+        if ("IntersectionObserver" in window && this.visibilityThreshold) {
             const observer = new IntersectionObserver(entries => {
                 entries.forEach(entry => {
                     if (!entry.isIntersecting) {
@@ -267,23 +259,49 @@ class VideoRTC extends HTMLElement {
                         this.connectedCallback();
                     }
                 });
-            }, {threshold: this.intersectionThreshold});
+            }, {threshold: this.visibilityThreshold});
             observer.observe(this);
         }
     }
 
-    internalWS() {
-        if (this.wsState !== WebSocket.CONNECTING) return;
-        if (this.ws) throw "connect with non null WebSocket";
+    /**
+     * Connect to WebSocket. Called automatically on `connectedCallback`.
+     * @return {boolean} true if the connection has started.
+     */
+    onconnect() {
+        if (!this.isConnected || !this.wsURL || this.ws || this.pc) return false;
 
-        const ts = Date.now();
+        // CLOSED or CONNECTING => CONNECTING
+        this.wsState = WebSocket.CONNECTING;
 
-        this.ws = new WebSocket(this.url);
+        this.connectTS = Date.now();
+
+        this.ws = new WebSocket(this.wsURL);
         this.ws.binaryType = "arraybuffer";
+        this.ws.addEventListener("open", ev => this.onopen(ev));
+        this.ws.addEventListener("close", ev => this.onclose(ev));
 
-        this.ws.addEventListener("open", () => {
-            console.debug("VideoRTC.ws.open", this.wsState);
+        return true;
+    }
 
+    ondisconnect() {
+        this.wsState = WebSocket.CLOSED;
+        if (this.ws) {
+            this.ws.close();
+            this.ws = null;
+        }
+
+        this.pcState = WebSocket.CLOSED;
+        if (this.pc) {
+            this.pc.close();
+            this.pc = null;
+        }
+    }
+
+    /**
+     * @returns {Array.<string>} of modes (mse, webrtc, etc.)
+     */
+    onopen() {
         // CONNECTING => OPEN
         this.wsState = WebSocket.OPEN;
 
@@ -301,58 +319,60 @@ class VideoRTC extends HTMLElement {
         this.ondata = null;
         this.onmessage = {};
 
-            let firstMode = "";
+        const modes = [];
 
         if (this.mode.indexOf("mse") >= 0 && "MediaSource" in window) { // iPhone
-                firstMode ||= "mse";
-                this.internalMSE();
+            modes.push("mse");
+            this.onmse();
         } else if (this.mode.indexOf("mp4") >= 0) {
-                firstMode ||= "mp4";
-                this.internalMP4();
+            modes.push("mp4");
+            this.onmp4();
         }
 
         if (this.mode.indexOf("webrtc") >= 0 && "RTCPeerConnection" in window) { // macOS Desktop app
-                firstMode ||= "webrtc";
-                this.internalRTC();
+            modes.push("webrtc");
+            this.onwebrtc();
         }
 
         if (this.mode.indexOf("mjpeg") >= 0) {
-                if (firstMode) {
+            if (modes.length) {
                 this.onmessage["mjpeg"] = msg => {
-                        if (msg.type !== "error" || msg.value.indexOf(firstMode) !== 0) return;
-                        this.internalMJPEG();
+                    if (msg.type !== "error" || msg.value.indexOf(modes[0]) !== 0) return;
+                    this.onmjpeg();
                 }
             } else {
-                    this.internalMJPEG();
+                modes.push("mjpeg");
+                this.onmjpeg();
             }
         }
-        });
 
-        this.ws.addEventListener("close", () => {
-            console.debug("VideoRTC.ws.close", this.wsState);
+        return modes;
+    }
 
-            if (this.wsState === WebSocket.CLOSED) return;
+    /**
+     * @return {boolean} true if reconnection has started.
+     */
+    onclose() {
+        if (this.wsState === WebSocket.CLOSED) return false;
 
         // CONNECTING, OPEN => CONNECTING
         this.wsState = WebSocket.CONNECTING;
         this.ws = null;
 
         // reconnect no more than once every X seconds
-            const delay = Math.max(this.RECONNECT_TIMEOUT - (Date.now() - ts), 0);
+        const delay = Math.max(this.RECONNECT_TIMEOUT - (Date.now() - this.connectTS), 0);
 
-            this.reconnectTimeout = setTimeout(() => {
-                this.reconnectTimeout = 0;
-                this.internalWS();
+        this.reconnectTID = setTimeout(() => {
+            this.reconnectTID = 0;
+            this.onconnect();
         }, delay);
-        });
+
+        return true;
     }
 
-    internalMSE() {
-        console.debug("VideoRTC.internalMSE");
-
+    onmse() {
         const ms = new MediaSource();
         ms.addEventListener("sourceopen", () => {
-            console.debug("VideoRTC.ms.sourceopen");
             URL.revokeObjectURL(this.video.src);
             this.send({type: "mse", value: this.codecs("mse")});
         }, {once: true});
@@ -361,9 +381,13 @@ class VideoRTC extends HTMLElement {
         this.video.srcObject = null;
         this.play();
 
+        this.mseCodecs = "";
+
         this.onmessage["mse"] = msg => {
             if (msg.type !== "mse") return;
 
+            this.mseCodecs = msg.value;
+
             const sb = ms.addSourceBuffer(msg.value);
             sb.mode = "segments"; // segments or sequence
             sb.addEventListener("updateend", () => {
@@ -372,10 +396,10 @@ class VideoRTC extends HTMLElement {
                     try {
                         sb.appendBuffer(buf.slice(0, bufLen));
                     } catch (e) {
-                        console.debug(e);
+                        // console.debug(e);
                     }
                     bufLen = 0;
-                } else if (sb.buffered.length) {
+                } else if (sb.buffered && sb.buffered.length) {
                     const end = sb.buffered.end(sb.buffered.length - 1) - 5;
                     const start = sb.buffered.start(0);
                     if (end > start) {
@@ -406,55 +430,12 @@ class VideoRTC extends HTMLElement {
         }
     }
 
-    internalRTC() {
-        console.debug("VideoRTC.internalRTC");
-
+    onwebrtc() {
         const pc = new RTCPeerConnection(this.pcConfig);
 
-        let mseCodecs = "";
-
         /** @type {HTMLVideoElement} */
         const video2 = document.createElement("video");
-        video2.addEventListener("loadeddata", () => {
-            console.debug("VideoRTC.video.loadeddata", video2.readyState, pc.connectionState);
-
-            if (pc.connectionState === "connected" || pc.connectionState === "connecting") {
-                // Video+Audio > Video, H265 > H264, Video > Audio, WebRTC > MSE
-                let rtcPriority = 0, msePriority = 0;
-
-                /** @type {MediaStream} */
-                const rtc = video2.srcObject;
-                if (rtc.getVideoTracks().length > 0) rtcPriority += 0x220;
-                if (rtc.getAudioTracks().length > 0) rtcPriority += 0x102;
-
-                if (mseCodecs.indexOf("hvc1.") >= 0) msePriority += 0x230;
-                if (mseCodecs.indexOf("avc1.") >= 0) msePriority += 0x210;
-                if (mseCodecs.indexOf("mp4a.") >= 0) msePriority += 0x101;
-
-                if (rtcPriority >= msePriority) {
-                    console.debug("VideoRTC.select RTC mode", rtcPriority, msePriority);
-
-                    this.video.controls = true;
-                    this.video.srcObject = rtc;
-                    this.play();
-
-                    this.pcState = WebSocket.OPEN;
-
-                    this.wsState = WebSocket.CLOSED;
-                    this.ws.close();
-                    this.ws = null;
-                } else {
-                    console.debug("VideoRTC.select MSE mode", rtcPriority, msePriority);
-
-                    pc.close();
-
-                    this.pcState = WebSocket.CLOSED;
-                    this.pc = null;
-                }
-            }
-
-            video2.srcObject = null;
-        }, {once: true});
+        video2.addEventListener("loadeddata", ev => this.onpcvideo(ev), {once: true});
 
         pc.addEventListener("icecandidate", ev => {
             const candidate = ev.candidate ? ev.candidate.toJSON().candidate : "";
@@ -462,8 +443,6 @@ class VideoRTC extends HTMLElement {
         });
 
         pc.addEventListener("track", ev => {
-            console.debug("VideoRTC.pc.track", ev.streams.length);
-
             // when stream already init
             if (video2.srcObject !== null) return;
 
@@ -477,30 +456,29 @@ class VideoRTC extends HTMLElement {
         });
 
         pc.addEventListener("connectionstatechange", () => {
-            console.debug("VideoRTC.pc.connectionstatechange", this.pc.connectionState);
-
             if (pc.connectionState === "failed" || pc.connectionState === "disconnected") {
                 pc.close(); // stop next events
 
                 this.pcState = WebSocket.CLOSED;
                 this.pc = null;
 
-                if (this.wsState === WebSocket.CLOSED && this.isConnected) {
-                    this.connectedCallback();
-                }
+                this.onconnect();
             }
         });
 
         this.onmessage["webrtc"] = msg => {
             switch (msg.type) {
                 case "webrtc/candidate":
-                    pc.addIceCandidate({candidate: msg.value, sdpMid: ""}).catch(() => console.debug);
+                    pc.addIceCandidate({
+                        candidate: msg.value,
+                        sdpMid: "0"
+                    }).catch(() => console.debug);
                     break;
                 case "webrtc/answer":
-                    pc.setRemoteDescription({type: "answer", sdp: msg.value}).catch(() => console.debug);
-                    break;
-                case "mse":
-                    mseCodecs = msg.value;
+                    pc.setRemoteDescription({
+                        type: "answer",
+                        sdp: msg.value
+                    }).catch(() => console.debug);
                     break;
                 case "error":
                     if (msg.value.indexOf("webrtc/offer") < 0) return;
@@ -522,9 +500,48 @@ class VideoRTC extends HTMLElement {
         this.pc = pc;
     }
 
-    internalMJPEG() {
-        console.debug("VideoRTC.internalMJPEG");
+    /**
+     * @param ev {Event}
+     */
+    onpcvideo(ev) {
+        /** @type {HTMLVideoElement} */
+        const video2 = ev.target;
+        const state = this.pc.connectionState;
 
+        // Firefox doesn't support pc.connectionState
+        if (state === "connected" || state === "connecting" || !state) {
+            // Video+Audio > Video, H265 > H264, Video > Audio, WebRTC > MSE
+            let rtcPriority = 0, msePriority = 0;
+
+            /** @type {MediaStream} */
+            const ms = video2.srcObject;
+            if (ms.getVideoTracks().length > 0) rtcPriority += 0x220;
+            if (ms.getAudioTracks().length > 0) rtcPriority += 0x102;
+
+            if (this.mseCodecs.indexOf("hvc1.") >= 0) msePriority += 0x230;
+            if (this.mseCodecs.indexOf("avc1.") >= 0) msePriority += 0x210;
+            if (this.mseCodecs.indexOf("mp4a.") >= 0) msePriority += 0x101;
+
+            if (rtcPriority >= msePriority) {
+                this.video.srcObject = ms;
+                this.play();
+
+                this.pcState = WebSocket.OPEN;
+
+                this.wsState = WebSocket.CLOSED;
+                this.ws.close();
+                this.ws = null;
+            } else {
+                this.pcState = WebSocket.CLOSED;
+                this.pc.close();
+                this.pc = null;
+            }
+        }
+
+        video2.srcObject = null;
+    }
+
+    onmjpeg() {
         this.ondata = data => {
             this.video.poster = "data:image/jpeg;base64," + VideoRTC.btoa(data);
         };
@@ -533,9 +550,7 @@ class VideoRTC extends HTMLElement {
         this.video.controls = false;
     }
 
-    internalMP4() {
-        console.debug("VideoRTC.internalMP4");
-
+    onmp4() {
         /** @type {HTMLVideoElement} */
         let video2;
 
@@ -580,5 +595,3 @@ class VideoRTC extends HTMLElement {
         return window.btoa(binary);
     }
 }
-
-customElements.define("video-rtc", VideoRTC);

www/video-stream.js

@@ -0,0 +1,85 @@
+import {VideoRTC} from "./video-rtc.js";
+
+class VideoStream extends VideoRTC {
+    constructor() {
+        super();
+
+        /** @type {HTMLDivElement} */
+        this.divMode = null;
+        /** @type {HTMLDivElement} */
+        this.divStatus = null;
+    }
+
+    /**
+     * Custom GUI
+     */
+    oninit() {
+        super.oninit();
+
+        this.innerHTML = `
+        <style>
+        .info {
+            position: absolute;
+            top: 0;
+            left: 0;
+            right: 0;
+            padding: 12px;
+            color: white;
+            display: flex;
+            justify-content: space-between;
+            pointer-events: none;
+        }
+        </style>
+        <div class="info">
+            <div class="status"></div>
+            <div class="mode"></div>
+        </div>
+        `;
+
+        this.divStatus = this.querySelector(".status");
+        this.divMode = this.querySelector(".mode");
+
+        const info = this.querySelector(".info")
+        this.insertBefore(this.video, info);
+    }
+
+    onconnect() {
+        const result = super.onconnect();
+        if (result) {
+            this.divMode.innerText = "loading";
+        }
+        return result;
+    }
+
+    onopen() {
+        const result = super.onopen();
+
+        this.onmessage["stream"] = msg => {
+            switch (msg.type) {
+                case "error":
+                    this.divMode.innerText = "error";
+                    this.divStatus.innerText = msg.value;
+                    break;
+                case "mse":
+                case "mp4":
+                case "mjpeg":
+                    this.divMode.innerText = msg.type.toUpperCase();
+                    this.divStatus.innerText = "";
+                    break;
+            }
+        }
+
+        return result;
+    }
+
+    onpcvideo(ev) {
+        super.onpcvideo(ev);
+
+        if (this.pcState !== WebSocket.CLOSED) {
+            this.divMode.innerText = "RTC";
+            this.divStatus.innerText = "";
+        }
+    }
+}
+
+customElements.define("video-stream", VideoStream);

www/webrtc.html

@@ -48,7 +48,7 @@
             console.debug('ws.onmessage', msg);
 
             if (msg.type === 'webrtc/candidate') {
-                pc.addIceCandidate({candidate: msg.value, sdpMid: ''});
+                pc.addIceCandidate({candidate: msg.value, sdpMid: '0'});
             } else if (msg.type === 'webrtc/answer') {
                 pc.setRemoteDescription({type: 'answer', sdp: msg.value});
             }