[wip] Docker Image for Media Streaming Expert User. Janus WebRTC Docker Image
Switch branches/tags
Nothing to show
Clone or download
Latest commit 8c86e1c Nov 11, 2018
Type Name Latest commit message Commit time
Failed to load latest commit information.
.travis.yml build May 5, 2017
Dockerfile 🐛 Nov 11, 2018
Makefile temporary fix Nov 29, 2017
README.md network bench Nov 10, 2018
load-test.png performance improvement Nov 10, 2018
system.png update system.png Oct 14, 2017


system Build Status CII Best Practices

Donation for transcoding developing

The reason for fundrasing is that developing transcording demo app is very heavy... But Many people want that!!!!

My bitcoin Address 3ESGqXHuaRqFCnAFSvBjJTfiqPc5Kn8xgc https://www.blockchain.com/btc/address/3ESGqXHuaRqFCnAFSvBjJTfiqPc5Kn8xgc?sort=0


  • janus docker image
  • janus performance improvement patch
  • jitsi vide bridge image ( in other repo)
  • example app for transcording
  • demo site for RTMP -> RTP -> WEBRTC
  • demo site for WEBRTC -> RTP -> RTMP
  • client video mixing in janus gateway
  • rtp => HLS with ffmpeg using GPU transcording


This is a docker image for Janus Webrtc Gateway. Janus Gateway is still under active development phase. So, as the official docs says, some minor modification of the middleware library versions happens frequently. I try to deal with such a chage as much as I can. If you need any request about this repo, free to contact me. About the details of setup for this docker image, you should read the official docs https://janus.conf.meetecho.com/index.html carefully.

Janus WebRTC Gateway Stability

tag v0.4.5 looks stable for my simple test. I will strongly recommend this version.

Janus WebRTC Gateway Performance

With the latest libnice, janus gateway seems to be great performance. This repo contains this patch(see https://gitlab.freedesktop.org/libnice/libnice/merge_requests/13 ) https://webrtchacks.com/sfu-load-testing/ load-test (right side janus graph is available for this docker image )

Janus WebRTC Gateway vs Jitsi Video Bridge(Personal Opinion)

I think that janus is better for webinar(web seminar), and jitsi is better for web conference system. The scalability of the current Jitsi Video Bridge(20181007) is poor because of having no local recording file(I'm not sure of this..). https://www.youtube.com/watch?v=OHHoqKCjJ0E Jitsi last-n + VP8 simulcasting has the very good performance for web conference https://jitsi.org/wp-content/uploads/2016/12/nossdav2015lastn.pdf For the video format, janus recording is per video streaming, jitsi is for mixed video conference by using chrome headlesss + ffmpeg(alsa, libxcb). From these points, janus is suitable for webinar, jitsi is for web conference. Of course, both WebRTC SFU are amazing work!! I'm using both.

[wip]Network benchmarking for preparing WebRTC SFU development

use iperf, netperf

Dockerfile Characteristics

janus ./configure

libsrtp version:           2.x
SSL/crypto library:        BoringSSL
DTLS set-timeout:          yes
DataChannels support:      no
Recordings post-processor: yes
TURN REST API client:      yes
Doxygen documentation:     no
    REST (HTTP/HTTPS):     yes
    WebSockets:            yes
    RabbitMQ:              no
    MQTT:                  no
    Unix Sockets:          no
    Nanomsg:               no
    Echo Test:             no
    Streaming:             yes
    Video Call:            no
    SIP Gateway (Sofia):   no
    SIP Gateway (libre):   no
    NoSIP (RTP Bridge):    no
    Audio Bridge:          yes
    Video Room:            yes
    Voice Mail:            no
    Record&Play:           no
    Text Room:             no
    Lua Interpreter:       no
    Duktape Interpreter:   no
Event handlers:
    Sample event handler:  no
    RabbitMQ event handler:no
    MQTT event handler:    no
JavaScript modules:        no


docker build --no-cache -t atyenoria/janus-gateway-docker .
docker run --rm --net=host --name="janus" -it -t atyenoria/janus-gateway-docker /bin/bash
docker exec -it /bin/bash janus

You should read the official doc https://janus.conf.meetecho.com/index.html carefully.


/root/bin/ffmpeg -y -i  "rtmp://$IP:80/rtmp_relay/$1  live=1"  -c:v libx264 -profile:v main -s 640x480  -an -preset ultrafast  -tune zerolatency -f rtp  rtp://$IP:$PORT

you should use janus streaming plugin


/root/bin/ffmpeg -analyzeduration 300M -probesize 300M -protocol_whitelist file,udp,rtp  -i $SDP_FILE  -c:v copy -c:a aac -ar 16k -ac 1 -preset ultrafast -tune zerolatency  -f flv rtmp://$IP:$PORT/rtmp_relay/atyenoria

In order to get the keyframe much easier, it is useful to set fir_freq=1 in janus conf
you should use janus video room or audiobridge plugin
After publishing your feed in your room, you should use rtp-forward. The sample javascript command is

# Input this in Google Chrome debug console. you must change publisher_id, room, video_port, host, secret for your conf.
var register = { "request" : "rtp_forward", "publisher_id": 3881836128186438, "room" : 1234, "video_port": 8050, "host" : "your ip address", "secret" : "unko" }
sfutest.send({"message": register});

[wip] Mixing for janus recording

  1. ffmpeg mixing from the janus recording outputs files I think that it is very difficult to align the file from the multiples timestamps in the case of the long mp4 file. you may consider the lipsync.
`#{ffmpeg_path} -y \
              -ss #{member[0].ss_at_time} -t #{member[0].t_at_time} -i #{member[0].file_path} -ss #{member[1].ss_at_time} -t #{member[1].t_at_time}  -i #{member[1].file_path} \
              -ss #{member[2].ss_at_time} -t #{member[2].t_at_time} -i #{member[2].file_path}  -f lavfi -i "color=White" \
                -filter_complex \"
                nullsrc=size=640x480 [base];
                [0:v] setpts=PTS-STARTPTS, scale=320x240 [upperleft];
                [1:v] setpts=PTS-STARTPTS, scale=320x240 [upperright];
                [2:v] setpts=PTS-STARTPTS, scale=320x240 [lowerleft];
                [3:v] setpts=PTS-STARTPTS, scale=320x240 [lowerright];
                [base][upperleft] overlay=shortest=1 [tmp1];
                [tmp1][upperright] overlay=shortest=1:x=320 [tmp2];
                [tmp2][lowerleft] overlay=shortest=1:y=240 [tmp3];
                [tmp3][lowerright] overlay=shortest=1:y=240:x=320;
                [0:a][1:a][2:a] amerge=inputs=3
              \" \
                -preset ultrafast -r 30 -b:v 300k -c:v libx264 #{"/tmp/" + @conference["room_name"] + "/" + index.to_s + ".mp4"}`
  1. jibri's solution headless chrome + grab the screen with ffmpeg is agressive approach. It is possible, but the scalabilitiy is poor. For example, the jibri's ffmpeg + chrome process consumes about 300% in my vps server.

Example nginx.conf for rtp => rtmp => hls for scalablity

server_names_hash_bucket_size 64;

server {
    listen 443 ssl;
    server_name temp;

    ssl_protocols TLSv1 TLSv1.1 TLSv1.2;
    ssl_prefer_server_ciphers on;

    add_header Strict-Transport-Security "max-age=31536000";

    ssl_certificate /usr/local/nginx/server.crt;
    ssl_certificate_key /usr/local/nginx/server.key;

    access_log  /app/log/nginx_access.log  ;
    error_log  /app/log/nginx_error.log  debug;

    location /janus {
         proxy_set_header X-Real-IP $remote_addr;
         proxy_set_header Host $host;
         proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
         proxy_set_header X-Forwarded-Proto $scheme;
         proxy_http_version 1.1;
         proxy_set_header Upgrade $http_upgrade;
         proxy_set_header Connection "upgrade";
         proxy_set_header Host $host;
         proxy_redirect off;


     location /janus_http {

    location /janus_admin {
         proxy_set_header X-Real-IP $remote_addr;
         proxy_set_header Host $host;
         proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
         proxy_http_version 1.1;
         proxy_set_header Upgrade $http_upgrade;
         proxy_set_header Connection "upgrade";
         proxy_set_header Host $host;

    location /janus_admin_http {

    location /hls {
         types {
             application/vnd.apple.mpegurl m3u8;
             video/mp2t ts;
         root /tmp;
         add_header Cache-Control no-cache;


Project Contributor

Akinori Nakajima https://twitter.com/atyenori

Anyone welcomed.