WebRTC-streamer

WebRTC-streamer is an experiment to stream video capture devices and RTSP sources through WebRTC using simple mechanism.

It embeds a HTTP server that implements API and serves a simple HTML page that use them through AJAX.

The WebRTC signaling is implemented through HTTP requests:

/api/call : send offer and get answer
/api/hangup : close a call
/api/addIceCandidate : add a candidate
/api/getIceCandidate : get the list of candidates

The list of HTTP API is available using /api/help.

Nowdays there is builds on CircleCI, CirrusCI and GitHub CI :

for x86_64 on Ubuntu Bionic
for armv7 crosscompiled (this build is running on Raspberry Pi2 and NanoPi NEO)
for armv6+vfp crosscompiled (this build is running on Raspberry PiB and should run on a Raspberry Zero)
for arm64 crosscompiled
Windows x64 build with clang
MacOS

The webrtc stream name could be :

an alias defined using -n argument then the corresponding -u argument will be used to create the capturer
an "rtsp://" url that will be openned using an RTSP capturer based on live555
an "file://" url that will be openned using an MKV capturer based on live555
an "screen://" url that will be openned by webrtc::DesktopCapturer::CreateScreenCapturer
an "window://" url that will be openned by webrtc::DesktopCapturer::CreateWindowCapturer
an "v4l2://" url that will capture H264 frames and store it using webrtc::VideoFrameBuffer::Type::kNative type (obviously not supported on Windows)
a capture device name

Dependencies :

It is based on :

WebRTC Native Code Package for WebRTC
civetweb HTTP server for HTTP server
live555 for RTSP/MKV source

Build

Install the Chromium depot tools (for WebRTC).

pushd ..
git clone https://chromium.googlesource.com/chromium/tools/depot_tools.git
export PATH=$PATH:`realpath depot_tools`
popd

Download WebRTC

mkdir ../webrtc
pushd ../webrtc
fetch --no-history webrtc 
popd

Build WebRTC Streamer

cmake . && make

It is possible to specify cmake parameters WEBRTCROOT & WEBRTCDESKTOPCAPTURE :

$WEBRTCROOT/src should contains source (default is $(pwd)/../webrtc)
WEBRTCDESKTOPCAPTURE enabling desktop capture if available (default is ON)

Usage

./webrtc-streamer [-H http port] [-S[embeded stun address]] -[v[v]]  [url1]...[urln]
./webrtc-streamer [-H http port] [-s[external stun address]] -[v[v]] [url1]...[urln]
./webrtc-streamer -V
	-v[v[v]]           : verbosity
	-V                 : print version
	-C config.json                     : load urls from JSON config file 
	-n name -u videourl -U audiourl    : register a name for a video url and an audio url
	[url]                              : url to register in the source list

	-H [hostname:]port : HTTP server binding (default 0.0.0.0:8000)
	-w webroot         : path to get files
	-c sslkeycert      : path to private key and certificate for HTTPS
	-N nbthreads       : number of threads for HTTP server
	-A passwd          : password file for HTTP server access
	-D authDomain      : authentication domain for HTTP server access (default:mydomain.com)

	-S[stun_address]                   : start embeded STUN server bind to address (default 0.0.0.0:3478)
	-s[stun_address]                   : use an external STUN server (default:stun.l.google.com:19302 , -:means no STUN)
	-t[username:password@]turn_address : use an external TURN relay server (default:disabled)
	-T[username:password@]turn_address : start embeded TURN server (default:disabled)
	-R [Udp port range min:max]        : Set the webrtc udp port range (default 0:65535)
	-W webrtc_trials_fields            : Set the webrtc trials fields (default:WebRTC-FrameDropper/Disabled/)		
	-a[audio layer]                    : spefify audio capture layer to use (default:0)		
	-q[filter]                         : spefify publish filter (default:.*)
	-o                                 : use null codec (keep frame encoded)

Arguments of '-H' are forwarded to option listening_ports of civetweb, then it is possible to use the civetweb syntax like -H8000,9000 or -H8080r,8443s.

Using -o allow to store compressed frame from backend stream using webrtc::VideoFrameBuffer::Type::kNative. This Hack the stucture webrtc::VideoFrameBuffer storing data in a override of i420 buffer. This allow to forward H264 frames from V4L2 device or RTSP stream to WebRTC stream. It use less CPU and have less adaptation (resize, codec, bandwidth are disabled).

Examples

./webrtc-streamer rtsp://wowzaec2demo.streamlock.net/vod/mp4:BigBuckBunny_115k.mov

Live Demo

We can access to the WebRTC stream using webrtcstreamer.html for instance :

An example displaying grid of WebRTC Streams is available using option "layout=x"

Live Demo

Using embedded STUN/TURN server behind a NAT:

It is possible start embeded ICE server and publish its url using:

./webrtc-streamer -S0.0.0.0:3478 -s$(curl -s ifconfig.me):3478
./webrtc-streamer -s- -T0.0.0.0:3478 -tturn:turn@$(curl -s ifconfig.me):3478
./webrtc-streamer -S0.0.0.0:3478 -s$(curl -s ifconfig.me):3478 -T0.0.0.0:3479 -tturn:turn@$(curl -s ifconfig.me):3479

The command curl -s ifconfig.me is getting the public IP, it could also given as a static parameter.

In order to configure the NAT rules using the upnp feature of the router, it is possible to use upnpc like this:

upnpc -r 8000 tcp 3478 tcp 3478 udp

Adapting with the HTTP port, STUN port, TURN port.

Embed in a HTML page:

Instead of using the internal HTTP server, it is easy to display a WebRTC stream in a HTML page served by another HTTP server. The URL of the webrtc-streamer to use should be given creating the WebRtcStreamer instance :

var webRtcServer      = new WebRtcStreamer(<video tag>, <webrtc-streamer url>);

A short sample HTML page using webrtc-streamer running locally on port 8000 :

<html>
<head>
<script src="libs/adapter.min.js" ></script>
<script src="webrtcstreamer.js" ></script>
<script>        
    var webRtcServer      = null;
    window.onload         = function() { 
        webRtcServer      = new WebRtcStreamer("video",location.protocol+"//"+window.location.hostname+":8000");
        webRtcServer.connect("rtsp://wowzaec2demo.streamlock.net/vod/mp4:BigBuckBunny_115k.mov");
    }
    window.onbeforeunload = function() { webRtcServer.disconnect(); }
</script>
</head>
<body> 
    <video id="video" />
</body>
</html>

Using WebComponent

Using web-component could be a simple way to display some webrtc stream, a minimal page could be :

<html>
<head>
    <script type="module" src="webrtc-streamer-element.js"></script>
</head>
<body>
   <webrtc-streamer url="rtsp://wowzaec2demo.streamlock.net/vod/mp4:BigBuckBunny_115k.mov"></webrtc-streamer>
</body>
</html>

Live Demo

Using the webcomponent with a stream selector :

Live Demo

Using the webcomponent over google map :

Live Demo

Object detection using tensorflow.js

Live Demo

Connect to Janus Gateway Video Room

A simple way to publish WebRTC stream to a Janus Gateway Video Room is to use the JanusVideoRoom interface

    var janus = new JanusVideoRoom(<janus url>, <webrtc-streamer url>)

A short sample to publish WebRTC streams to Janus Video Room could be :

<html>
<head>
<script src="janusvideoroom.js" ></script>
<script>        
	var janus = new JanusVideoRoom("https://janus.conf.meetecho.com/janus", null);
	janus.join(1234, "rtsp://pi2.local:8554/unicast","pi2");
	janus.join(1234, "rtsp://wowzaec2demo.streamlock.net/vod/mp4:BigBuckBunny_115k.mov","media");	    
</script>
</head>
</html>

Live Demo

This way the communication between Janus API and WebRTC Streamer API is implemented in Javascript running in browser.

The same logic could be implemented in NodeJS using the same JS API :

global.request = require('then-request');
var JanusVideoRoom = require('./html/janusvideoroom.js'); 
var janus = new JanusVideoRoom("http://192.168.0.15:8088/janus", "http://192.168.0.15:8000")
janus.join(1234,"mmal service 16.1","video")

Connect to Jitsi

A simple way to publish WebRTC stream to a Jitsi Video Room is to use the XMPPVideoRoom interface

    var xmpp = new XMPPVideoRoom(<xmpp server url>, <webrtc-streamer url>)

A short sample to publish WebRTC streams to a Jitsi Video Room could be :

<html>
<head>
<script src="libs/strophe.min.js" ></script>
<script src="libs/strophe.muc.min.js" ></script>
<script src="libs/strophe.disco.min.js" ></script>
<script src="libs/strophe.jingle.sdp.js"></script>
<script src="libs/jquery-3.5.1.min.js"></script>
<script src="xmppvideoroom.js" ></script>
<script>        
	var xmpp = new XMPPVideoRoom("meet.jit.si", null);
	xmpp.join("testroom", "rtsp://wowzaec2demo.streamlock.net/vod/mp4:BigBuckBunny_115k.mov","Bunny");	    
</script>
</head>
</html>

Live Demo

Docker image

You can start the application using the docker image :

    docker run -p 8000:8000 -it mpromonet/webrtc-streamer

You can expose V4L2 devices from your host using :

    docker run --device=/dev/video0 -p 8000:8000 -it mpromonet/webrtc-streamer

The container entry point is the webrtc-streamer application, then you can :

get the help using :

  docker run -p 8000:8000 -it mpromonet/webrtc-streamer -h

run the container registering a RTSP url using :

  docker run -p 8000:8000 -it mpromonet/webrtc-streamer -n raspicam -u rtsp://pi2.local:8554/unicast

run the container giving config.json file using :

  docker run -p 8000:8000 -v $PWD/config.json:/app/config.json mpromonet/webrtc-streamer

entheosl / webrtc-streamer

WebRTC-streamer

Dependencies :

Build

Install the Chromium depot tools (for WebRTC).

Download WebRTC

Build WebRTC Streamer

Usage

Examples

Using embedded STUN/TURN server behind a NAT:

Embed in a HTML page:

Using WebComponent

Object detection using tensorflow.js

Connect to Janus Gateway Video Room

Connect to Jitsi

Docker image

About

Languages