from Hacker News

My video streaming wishlist for the next 3 to 5 years

by programjoe on 11/30/23, 2:53 PM with 5 comments

I work in the video streaming space and curious on others thoughts. Here's my wishlist though

1. WebTransport and WebCodec becomes the primary means for client to server real-time video delivery (e.g. compositing, off-device analysis)

2. No more vendor lockin with WebRTC (WHIP and WHEP might help here). Build a solution once on the client and if I don't like my provider just change the endpoint URL.

3. Google MediaPipe or a high level API on the browser to run AI models easily for audio / video. Right now it seems like most solutions for simple things like blurring are just minor abstractions on top of MediaStreamTrackProcessor.

4. Optimized headless browser for cloud rendering. Too many terrible solutions at the moment using CEF and chrome that then use ffmpeg or gstreamer, XVFB and pulseaudio.

5. Plug and play pipelines in the cloud for video processing (like zapier for video). I can plug in any processing I want in between the source and sink without a convulted mess of trying to push audio and video around to different apps either in network or across the internet.

  • by toomuchtodo on 11/30/23, 3:03 PM

    Hey! Just a rando here, but I would be interested in hearing your opinion as to where Peertube does well with this wish list and where it needs improvement.

    https://joinpeertube.org

    https://framablog.org/2023/11/28/peertube-v6-is-out-and-powe...

  • by Sean-Der on 11/30/23, 3:03 PM

    I see 1 and 2 as going in the opposite directions. WebTransport+WebCodec enables the shipping of binary blobs for each individual service. WHIP+WHEP might see enough demand (OBS input) that locked down services have to offer it.

    What cloud rendering are you trying to do? My hope/goal is to drop the browser dependency completely.