I’ve used the simple AIORTC webcam example to make an WebRTC server from an RTSP flow which can be queried from any browser:

async def offer(request):
    ...
    peer_connection = RTCPeerConnection()
    peer_connections.add(peer_connection)
    ...
    # HERE! One player is created for each WebRTC OFFER
    player = MediaPlayer("rtsp://myserver:8554/live.sdp")

    video = MediaRelay().subscribe(player.video)
    peer_connection.addTrack(video)
    ...

The problem here is that I need multiple clients to connect to the server. When multiple (browser) clients are connected, there’s a huge memory consumption.

An evident solution to consume less memory is to create a single player for all browser clients:

# HERE! A single player is created for all WebRTC OFFERs
player = MediaPlayer("rtsp://myserver:8554/live.sdp")
async def offer(request):
    ...
    peer_connection = RTCPeerConnection()
    peer_connections.add(peer_connection)
    ...
    # Will use the unique player
    video = MediaRelay().subscribe(player.video)
    peer_connection.addTrack(video)
    ...

This solution is great from the memory pov (80Mb for the application + 100Mb for each client). However, there is a HUGE problem: clients receive interleaved frames!

I imagine that each client calls the .recv() method, so if there are two clients, client 1 gets frames 1,3,5… and client 2 gets frames 2,4,6…

My knowledge of webrtc/video is very limited, is there any simple solution for this interleaving, or should I overload the MediaPlayer so to overload the PlayerStream.recv() method… (seems very complex!)?