This blog is in continuation to the attempts / outcomes and problems in building a WebRTC to RTP media framework that successfully stream / broadcast WebRTC content to non webrtc supported browsers ( safari / IE ) / media players ( VLC ).
Attempt 4: Stream the content to a WebRTC endpoint which is hidden in a video call . Pick the stream from vp8 object URL send to a streaming server
This process involved the following components :
WebRTC API : simplewebrtc on Chrome
Transfer mechanism from client to Streaming server: webrtc media channel
Problems : No streaming server is qualified to handle a direct webrtc input and stream it on network .
Attempt 4.1 : Stream the content to a WebRTC endpoint . Do WebRTC Endpoint to RTP Endpoint bridge using Kurento APIs.
Use the RTP port and ip address to input into a ffmpeg or gstreamer or VLC terminal command and out put a live H264 stream on another ip and port address .
This process involved the following components :
API : Kurento
Transfer mechanism : HTML5 webrtc client -> application server hosting java -> media server -> application for webrtc media to RTP media conversation -> RTP player
Screenshots of attempts with Wowza to stream RTP from a IP and port
Problems : The stream was black which means 100% loss.
Lesson learned : RTP is not suitable for over the intgernet transmission especially with firewalls
Attempt 4.2 : Build a WebRTC Endpoint to Http endpoint in kurento and force the video audio encoding to be that of H264 and PCMU.
Code snippet for adding constraints to output media via pipeline and forcing choice of codecs( H264 for video and PCMU for audio ).
MediaPipeline pipeline = kurento.createMediaPipeline();
WebRtcEndpoint webRtcEndpoint = new WebRtcEndpoint.Builder(pipeline).build();
HttpGetEndpoint httpEndpoint=new HttpGetEndpoint.Builder(pipeline).build();
org.kurento.client.Fraction fr= new org.kurento.client.Fraction(1, 30);
VideoCaps vc= new VideoCaps(VideoCodec.H264,fr);
httpEndpoint.setVideoFormat(vc);
AudioCaps ac= new AudioCaps(AudioCodec.PCMU, 65536);
httpEndpoint.setAudioFormat(ac);
webRtcEndpoint.connect(httpEndpoint);
Alternatively one can opt to use gstreamer filter to force the output in raw format.
// basic media operation of 1 pipeline and 2 endpoints
MediaPipeline pipeline = kurento.createMediaPipeline();
WebRtcEndpoint webRtcEndpoint = new WebRtcEndpoint.Builder(pipeline).build();
RtpEndpoint rtpEndpoint = new RtpEndpoint.Builder(pipeline).build();
// adding Gstream filters
GStreamerFilter filter1 = new GStreamerFilter.Builder(pipeline, "videorate max-rate=30").withFilterType(FilterType.VIDEO).build();
GStreamerFilter filter2 = new GStreamerFilter.Builder(pipeline, "capsfilter caps=video/x-h264,width=1280,height=720,framerate=30/1").withFilterType(FilterType.VIDEO).build();
GStreamerFilter filter3 = new GStreamerFilter.Builder(pipeline, "capsfilter caps=audio/x-mpeg,layer=3,rate=48000").withFilterType(FilterType.AUDIO).build();
// connecting all poin ts to one another
webRtcEndpoint.connect (filter1);
filter1.connect (filter2);
filter2.connect (filter3);
filter3.connect (rtpEndpoint);
// RTP SDP offer and answer
String requestRTPsdp = rtpEndpoint.generateOffer();
rtpEndpoint.processAnswer(requestRTPsdp);
End result : The output is still webm based and doesnt work on h264 clients.
Attempt 5 : Use a RTP SDP Endpoint ( ie a SDP file valid for a given session ) and use it to play the WebRTC media over Wowza streaming server
This process involved the following components
WebRTC Stream and object URL of the blob containing VP8 media
Kurento WebRTC Endpoint bridge to generate SDP
Wowza Streaming server
Snippet used for kurento to generate a SDP file from WebRTC to RTP bridge
@RequestMapping(value = "/rtpsdp", method = RequestMethod.POST)
private String processRequestrtpsdp(@RequestBody String sdpOffer)
throws IOException, URISyntaxException, InterruptedException {
//basic media operation of 1 pipeline and 2 endpoinst
MediaPipeline pipeline = kurento.createMediaPipeline();
WebRtcEndpoint webRtcEndpoint = new WebRtcEndpoint.Builder(pipeline).build();
RtpEndpoint rtpEndpoint = new RtpEndpoint.Builder(pipeline).build();
//connecting all poin ts to one another
webRtcEndpoint.connect (rtpEndpoint);
// RTP SDP offer and answer
String requestRTPsdp = rtpEndpoint.generateOffer();
rtpEndpoint.processAnswer(requestRTPsdp);
// write the SDP conector to an external file
PrintWriter out = new PrintWriter("/tmp/test.sdp");
out.println(requestRTPsdp);
out.close();
HttpGetEndpoint httpEndpoint = new HttpGetEndpoint.Builder(pipeline).build();
PlayerEndpoint player = new PlayerEndpoint.Builder(pipeline, requestRTPsdp).build();
httpEndpoint.connect(rtpEndpoint);
player.connect(httpEndpoint);
// Playing media and opening the default desktop browser
player.play();
String videoUrl = httpEndpoint.getUrl();
System.out.println(" ------- video URL -------------"+ videoUrl);
// send the response to front client
String responseSdp = webRtcEndpoint.processOffer(sdpOffer);
return responseSdp;
}
End result : wowza doesnt not recognize the WebRTC SDP and play the video
screenshot of wowza with SDP input
Attempt 5.1 : Use a RTP SDP Endpoint ( ie a SDP file valid for a given session ) and use it to play the WebRTC media over Default Ubuntu media player
End result : wowza doesnt not recognize the WebRTC SDP and play the video : deformed media
screenshot of playing from a SDP file
Attempt 5.2 : Use a RTP SDP Endpoint ( ie a SDP file valid for a given session ) and use it to play the WebRTC media over VLC using socket input
End result : nothing plays
screenshot of VLC connected to play from socket and failure to play anything
Attempt 5.3: Create a WebRTC endpoint and connected it to RTP endpoint via media pipelines . Also make the RTP SDP offer and answering the same . Play with ffnpeg / ffplay / gst playbin
Write the requestRTPsdp to a file and obtain a RTP connector endpoint with Application/SDP .It plays okay with gst playbin ( 10 secs without audio ). Successful attempt to play from a gst playbin
gst-launch -vvv playbin uri=file:///tmp/test.sdp
but refuses to be played by VLC , ffplay and even wowza . The error generated with
End result : This results in “Could not find codec parameter for stream1 ( video:h263, none ) .Other errors types are , Could not write header for output file output file is empty nothing was encoded”
Error screenshots of trying to play the RTP SDP file with ffmpeg
Attempt 6 : Use a WebRTC capable media and streaming server ( eg Kurento ) to pick a live stream of VP8 .
Convert the VP8 to H264 ( ffmpeg / RTP endpoint )
Convert H264 to Mp4 using MP4 parser and pass to a streaming server ( wowza)
End Result : yes it did work on mozilla but with considerable lag
Update : Thankfully the updates to WebRTC standards mandated the support for PCMU and AVC/H264 CB profile in the media stack of the browser thus solving the “from scratch buildup of transcoder between webrtc and non webrtc endpoints”.
Video Codecs : RFC 7742 specifies that all WebRTC-compatible browsers must support VP8 and H.264’s Constrained Baseline profile for video.
Audio Codecs : RFC 7874 specifies that browsers must support at least the Opus codec as well as G.711’s PCMA and PCMU formats.
The latest Webrtc specification lists a set of codecs which all compliant browsers are required to support which includes chrome 52 , Firefox , safari , edge.
References :
RFC7742: WebRTC Video Processing and Codec Requirements
RFC 7874: WebRTC Audio Codec and Processing Requirements
As the title of this article suggests I am going to pen my attempts of streaming / broadcasting Live Video WebRTC call to non WebRTC supported browsers and media players such as VLC , ffplay , default video player in Linux etc.
Some of the high level archietctures for streaming Webrtc Video to multiple endpoints can be viewed in the post below.
Aim : I will be attempting to create a lightweight WebRTC to raw/h264 transcoder by making my own media engine which takes input from WebRTC peerconnection or getusermedia. I am sharing my past experiments in hope of helping someone whose objective may be to acheive the same since many non webrtc supported endpoints ( Rpi , kisosks , mobile browsers ) could benifit heavily from webrtc streaming . Even if your objective is not the same as mine, you may gain some insigh on what not to do when making a media transcoder.
It uses API fromwebrtc-experiment.com. The broadcast is in one direction only where the viewrs are never asked for their mic / webcam permission .
problem : The broadcast is for WebRTC browsers only and doesnt support non webrtc players / browsers
Attempt 1.1: Stream the media directly to nodejs through websocke
window.addEventListener('DOMContentLoaded', function () {
var v = document.getElementById('v');
navigator.getUserMedia = (navigator.getUserMedia ||
navigator.webkitGetUserMedia ||
navigator.mozGetUserMedia ||
navigator.msGetUserMedia);
if (navigator.getUserMedia) {
// Request access to video only
navigator.getUserMedia(
{
video: true,
audio: false
},
function (stream) {
var url = window.URL || window.webkitURL;
v.src = url ? url.createObjectURL(stream) : stream;
v.play();
var ws = new WebSocket('ws://localhost:3000', 'echo-protocol');
waitForSocketConnection(ws, function () {
console.log(" url.createObjectURL(stream)-----", url.createObjectURL(stream))
ws.send(stream);
console.log("message sent!!!");
});
},
function (error) {
alert('Something went wrong. (error code ' + error.code + ')');
return;
}
);
} else {
alert('Sorry, the browser you are using doesn\'t support getUserMedia');
return;
}
});
//Make the function wait until the connection is made...
function waitForSocketConnection(socket, callback) {
setTimeout(
function () {
if (socket.readyState === 1) {
console.log("Connection is made")
if (callback != null) {
callback();
}
return;
} else {
console.log("wait for connection...")
waitForSocketConnection(socket, callback);
}
}, 5); // wait 5 milisecond for the connection...
}
Problem : The video is in form of buffer and doesnot play
Attempt 2: Record the WebRTC media ( 5 secs each ) into chunks of webm format-> transfer them to other end -> append the chunks together like a regular file
This process involved the following components :
Recorder Javascript library : RecordJs
Transfer mechanism : Record using RecordRTC.js -> send to other end for media server -> stitching together the small webm files into big one at runtime and play
Programs :
Code for video recorder
navigator.getUserMedia(videoConstraints, function (stream) {
video.onloadedmetadata = function () {
video.width = 320;
video.height = 240;
var options = {
type: isRecordVideo ? 'video' : 'gif',
video: video,
canvas: {
width: canvasWidth_input.value,
height: canvasHeight_input.value
}
};
recorder = window.RecordRTC(stream, options);
recorder.startRecording();
};
video.src = URL.createObjectURL(stream);
}, function () {
if (document.getElementById('record-screen').checked) {
if (location.protocol === 'http:')
alert('https is mandatory to capture screen.');
else
alert('Multi-capturing of screen is not allowed.Have you enabled flag: "Enable screen capture support in getUserMedia"?');
} else
alert('Webcam access is denied.');
});
Code for video append-er
var FILE1 = '1.webm';
var FILE2 = '2.webm';
var FILE3 = '3.webm';
var FILE4 = '4.webm';
var FILE5 = '5.webm';
var NUM_CHUNKS = 5;
var video = document.querySelector('video');
window.MediaSource = window.MediaSource || window.WebKitMediaSource;
if (!!!window.MediaSource) {
alert('MediaSource API is not available');
}
var mediaSource = new MediaSource();
video.src = window.URL.createObjectURL(mediaSource);
function callback(e) {
var sourceBuffer = mediaSource.addSourceBuffer('video/webm; codecs="vorbis,vp8"');
GET(FILE1, function (uInt8Array) {
var file = new Blob([uInt8Array], {type: 'video/webm'});
var i = 1;
(function readChunk_(i) {
var reader = new FileReader();
reader.onload = function (e) {
sourceBuffer.appendBuffer(new Uint8Array(e.target.result));
if (i == NUM_CHUNKS) mediaSource.endOfStream();
else {
if (video.paused) {
video.play(); // Start playing after 1st chunk is appended.
}
readChunk_(++i);
}
};
reader.readAsArrayBuffer(file);
})(i); // Start the recursive call by self calling.
});
}
mediaSource.addEventListener('sourceopen', callback, false);
mediaSource.addEventListener('webkitsourceopen', callback, false);
mediaSource.addEventListener('webkitsourceended', function (e) {
logger.log('mediaSource readyState: ' + this.readyState);
}, false);
// function get the video via XHR
function GET(url, callback) {
var xhr = new XMLHttpRequest();
xhr.open('GET', url, true);
xhr.responseType = 'arraybuffer';
xhr.send();
xhr.onload = function (e) {
if (xhr.status != 200) {
alert("Unexpected status code " + xhr.status + " for " + url);
return false;
}
callback(new Uint8Array(xhr.response));
};
}
Shortcoming of this approach
The webm files failed to play on most of the media players
The recorder can only either record video or audio file at a time .
Attempt 2.Chunking and media proxy
Since the previous approach failed to support on webrtc endpoinst , the next iteration of this approach was to channel the webrtc media via a nodejs server thus disrupting the peer to peer media strem in favour of centralized / proxied emdia stream. This would enable me to obtain raw media packets form teh stream using low level C based vp8 decoder libraries and then re encode them to h364 or other media formats suitable for endpoints .
In theory media could be reencoded jusing openH264 library and the frame could be then send to players
let mediaSource = new MediaSource();
let sourceBuffer = mediaSource.addSourceBuffer('video/mp4; codecs=vp9',
new VP9Decoder());
let buffer = await loadBuffer();
sourceBuffer.appendBuffer(buffer);
Further extending for uncompressed video
let mediaSource = new MediaSource();
let sourceBuffer = mediaSource.addSourceBuffer('video/raw; codecs=yuv420p');
for (let p in demuxPAckets()) {
let frame = await codec.decode(p);
sourceBuffer.appendBuffer(frame);
}
Atleast that was the plan .
Attempt 2.1: Record the WebRTC media ( 5 secs each ) into chunks of webm format ( RecordRTC.js) > Use Kurento JS script ( kws-media-api,js) to make a HTTP Endpoint to recorded Webm files -> append the chunks together like a regular file at runtime
// UI elements
function getByID(id) {
return document.getElementById(id);
}
var recordAudio = getByID('record-audio'),
recordVideo = getByID('record-video'),
stopRecordingAudio = getByID('stop-recording-audio'),
stopRecordingVideo = getByID('stop-recording-video'),
broadcasting = getByID('broadcasting');
var canvasWidth_input = getByID('canvas-width-input'),
canvasHeight_input = getByID('canvas-height-input');
var video = getByID('video');
var audio = getByID('audio');
// Audio video constraints
var videoConstraints = {
audio: false,
video: {
mandatory: {},
optional: []
}
};
var audioConstraints = {
audio: true,
video: false
};
// Recording and stop recording - to be convrted into real time capture and chunking
const ws_uri = 'ws://localhost:8888/kurento';
var URL_SMALL = "http://localhost:8080/streamtomp4/approach1/5561840332.webm";
var audioStream;
var recorder;
recordAudio.onclick = function () {
if (!audioStream)
navigator.getUserMedia(audioConstraints, function (stream) {
if (window.IsChrome) stream = new window.MediaStream(stream.getAudioTracks());
audioStream = stream;
audio.src = URL.createObjectURL(audioStream);
audio.muted = true;
audio.play();
// "audio" is a default type
recorder = window.RecordRTC(stream, {
type: 'audio'
});
recorder.startRecording();
}, function () {
});
else {
audio.src = URL.createObjectURL(audioStream);
audio.muted = true;
audio.play();
if (recorder) recorder.startRecording();
}
window.isAudio = true;
this.disabled = true;
stopRecordingAudio.disabled = false;
};
Recording and stop recording video inot small media files ( chunks )
recordVideo.onclick = function () {
recordVideoOrGIF(true);
};
stopRecordingAudio.onclick = function () {
this.disabled = true;
recordAudio.disabled = false;
audio.src = '';
if (recorder)
recorder.stopRecording(function (url) {
audio.src = url;
audio.muted = false;
audio.play();
document.getElementById('audio-url-preview').innerHTML = '<a href="' + url + '" target="_blank">Recorded Audio URL</a>';
});
};
function recordVideoOrGIF(isRecordVideo) {
navigator.getUserMedia(videoConstraints, function (stream) {
video.onloadedmetadata = function () {
video.width = 320;
video.height = 240;
var options = {
type: isRecordVideo ? 'video' : 'gif',
video: video,
canvas: {
width: canvasWidth_input.value,
height: canvasHeight_input.value
}
};
recorder = window.RecordRTC(stream, options);
recorder.startRecording();
};
video.src = URL.createObjectURL(stream);
}, function () {
if (document.getElementById('record-screen').checked) {
if (location.protocol === 'http:')
alert('<https> is mandatory to capture screen.');
else
alert('Multi-capturing of screen is not allowed. Capturing process is denied. Are you enabled flag: "Enable screen capture support in getUserMedia"?');
} else
alert('Webcam access is denied.');
});
window.isAudio = false;
if (isRecordVideo) {
recordVideo.disabled = true;
stopRecordingVideo.disabled = false;
} else {
recordGIF.disabled = true;
stopRecordingGIF.disabled = false;
}
}
stopRecordingVideo.onclick = function () {
this.disabled = true;
recordVideo.disabled = false;
if (recorder)
recorder.stopRecording(function (url) {
video.src = url;
video.play();
document.getElementById('video-url-preview').innerHTML = '<a href="' + url + '" target="_blank">Recorded Video URL</a>';
});
};
Broadcasting the chunks to media engine
function onerror(error) {
console.log(" error occured");
console.error(error);
}
broadcast.onclick = function () {
var videoOutput = document.getElementById("videoOutput");
KwsMedia(ws_uri, function (error, kwsMedia) {
if (error) return onerror(error);
// Create pipeline
kwsMedia.create('MediaPipeline', function (error, pipeline) {
if (error) return onerror(error);
// Create pipeline media elements (endpoints & filters)
pipeline.create('PlayerEndpoint', {uri: URL_SMALL}, function (error, player) {
if (error) return console.error(error);
pipeline.create('HttpGetEndpoint', function (error, httpGet) {
if (error) return onerror(error);
// Connect media element between them
player.connect(httpGet, function (error, pipeline) {
if (error) return onerror(error);
// Set the video on the video tag
httpGet.getUrl(function (error, url) {
if (error) return onerror(error);
videoOutput.src = url;
console.log(url);
// Start player
player.play(function (error) {
if (error) return onerror(error);
console.log('player.play');
});
});
});
// Subscribe to HttpGetEndpoint EOS event
httpGet.on('EndOfStream', function (event) {
console.log("EndOfStream event:", event);
});
});
});
});
}, onerror);
}
problem : dissecting the live video into small the files and appending to each other on reception is an expensive , time and resource consuming process . Also involves heavy buffering and other problems pertaining to real-time streaming .
Attempt 2.2 : Send the recorded chunks of webm to a port on linux server. Use socket programming to pick up these individual files and play using VLC player from UDP port of the Linux Server
End Result : Small file containers play but slow buffering makes this approach non compatible for streaming files chunks and appending as single file.
Attempt 2.3: Send the recorded chunks of webm to a port on linux server socket . Use socket programming to pick up these individual webm files and convert to H264 format so that they can be send to a media server.
This process involved the following components :
Recorder Javascript library : RecordJs
Transfer mechanism :WebRTC endpoint -> Call handler ( Record in chunks ) -> ffmpeg / gstreamer to put it on RTP -> streaming server like wowza – > viewers
Programs : Use HTML webpage Webscoket connection -> nodejs program to write content from websocket to linux socket -> nodejs program to read that socket and print the content on console
Snippet to transfer the webm recorder files over websocket to nodejs program
// Make the function wait until the connection is made.
function waitForSocketConnection(socket, callback) {
setTimeout(
function () {
if (socket.readyState === 1) {
console.log("Connection is made")
if (callback != null)
callback();
} else {
console.log("wait for connection...")
waitForSocketConnection(socket, callback);
}
}, 5); // wait 5 milisecond for the connection...
}
function previewFile() {
var preview = document.querySelector('img');
var file = document.querySelector('input[type=file]').files[0];
var reader = new FileReader();
reader.onloadend = function () {
preview.src = reader.result;
console.log(" reader result ", reader.result);
var video = document.getElementById("v");
video.src = reader.result;
console.log(" video played ");
var ws = new WebSocket('ws://localhost:3000', 'echo-protocol');
waitForSocketConnection(ws, function () {
ws.send(reader.result);
console.log("message sent!!!");
});
}
if (file) {
// converts to base64 encoded string of the file data
//reader.readAsDataURL(file);
reader.readAsBinaryString(file);
} else {
preview.src = "";
}
}
Program for Linux Sockets sender which creates the socket for the webm files in nodejs
var net = require('net');
var fs = require('fs');
var socketPath = '/tmp/tfxsocket';
var http = require('http');
var stream = require('stream');
var util = require('util');
var WebSocketServer = require('ws').Server;
var port = 3000;
var serverUrl = "localhost";
var socket;
/*----------http server -----------*/
var server = http.createServer(function (request, response) {});
server.listen(port, serverUrl);
console.log('HTTP Server running at ', serverUrl, port);
/*------websocket server ----------*/
var wss = new WebSocketServer({server: server});
wss.on("connection", function (ws) {
console.log("websocket connection open");
ws.on('message', function (message) {
console.log(" stream recived from broadcast client on port 3000 ");
var s = require('net').Socket();
s.connect(socketPath);
s.write(message);
console.log(" send the stream to socketPath", socketPath);
});
ws.on("close", function () {
console.log("websocket connection close")
});
});
Program for Linux Socket Listener using nodejs and socket . Here the socket is in node /tmp/mysocket
var net = require('net');
var client = net.createConnection("/tmp/mysocket");
client.on("connect", function() {
console.log("connected to mysocket");
});
client.on("data", function(data) {
console.log(data);
});
client.on('end', function() {
console.log('server disconnected');
});
Output 1: Video Buffer displayed
Output 2 : Payload from Video displayed that shows the pipeline works but no output yet.
ffmpeg format of transfering the content from socket to UDP IP and port
ffmpeg -i unix://tmp/mysocket -f format udp://192.168.0.119:8083
problems of this approach : The video was on a passing stage from the socket and contained no information as such when tried to play / show console.
Attempt 3 : Use existing media engine like kurento to do the transocding for me
Send the live WebRTC stream from Kurento WebRTC endpoint to Kurento HTTP endpoint then play using Mozilla VLC web plugin