Skip to content

Small but complete example of how to use WebRTC to setup voice chat between 2+ people.

License

Notifications You must be signed in to change notification settings

vigilancer/webrtc-group-chat-example

 
 

Repository files navigation

About this fork

This fork is intended for use only with audio so video support was removed because I have no intentions to test and support it. Also I don't have a webcam.

Second, HTTPS support was added because without it this demo was working for Firefox users only (which allows getUserMedia for insecure hosts). Now support should be much broader, tested Chrome and Firefox on Mac, Chrome on Android. From now on client.html and signaling server should be hosted on SSL enabled domain.

3rd, some issues were fixed (but some deprecation issues are still there).

PS: hosting slightly modified version of this with channels support. feel free to explore: https://ham.ayeaye.ae

About

This is a "simple", but complete example of how to utilize WebRTC to do peer to peer voice chatting between two or more people.

Server Side

This example uses node.js and socket.io to create a "Signaling Server", which runs on (or near) your web server to manage who should talk to who. The purpose of the signaling server is to relay information between peers while you are setting them up to talk directly to each other.

Client Side

Included is client.html which contains all of the logic to connect to the signaling server, join a virtual group chat channel, connect with peers, and stream audio to all party members using the raw WebRTC API.

Running

Node.js signaling server

You'll need to install node.js as well as the express and socket.io libraries:

npm install

Then change in signaling-server.js variables SSL_KEY_PATH and SSL_CERT_PATH to the full path to key and certificate of domain you are hosting this demo on.

And then simply run the signaling server:

node signaling-server.js

Web server

You'll also probably want to host client.html on a web server somewhere. You'll need to edit client.html and change YOUR_HTTPS_DOMAIN to be the hostname or IP of the signaling server.

Note: you cannot file locally or AFAIK host signaling server and client.html on different hosts.

Running the sample

Now navigate to wherever you stuck client.html and you should be presented with a dialog asking permission to access your microphone / webcam. Once accepted, you should see a local stream appear on the page. Now open up the same page in another browser, on the same computer or another, and watch as the magic of WebRTC takes effect and both images and audio samples mysteriously move from one browser to the other. Repeat with as many browsers as you dare.

Note: At the time of writing this, only firefox and chrome support WebRTC, however both browsers support this on Windows, Linux, Mac, and Android, so lots of fun can be had pointing everyones' phones, tablets, and laptops at that client.html and bogging down your network with audio traffic.

Using things other than jQuery, node.js, and socket.io

The choice of node.js and socket.io is based purely on my familiarity with them and the fact that their fairly easy to understand even if you aren't familiar with them. However, you can use any mechanisms you want for your signaling system, you just need a way to exchange ICE candidates and session descriptions between clients.

The use of jQuery is even less important, I just like using it for DOM manipulation, and we only do that to add and remove the elements in this demo. We don't use it at all for anything WebRTC specific in this example.

Adapter.js

You'll see client.html use adapter.js. This "library" just normalizes the WebRTC API, which will only be necessary while WebRTC is making its way through the standards process. Once everything is standardized and functions are de-prefixed in the browsers, this won't be necessary anymore.

About

Small but complete example of how to use WebRTC to setup voice chat between 2+ people.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • HTML 77.0%
  • JavaScript 22.8%
  • Shell 0.2%