Neural codec bitstreaming POC

[johndpope]

just looking for some high level insights -
i skimmed in the paper that you were doing testing with IMF bit streaming....
what language did you build that out in? rust? wasm? c++ or??

if i was to build a codec for academic purposes - is your first call to go web?

there seems to be demand for good streaming in this space - and the IMF model seems like SOTA.
you didn't attempt to go 1024x1024 ? would it mean doubling the m latent -> 32 -> 64?

Im confused how the model would be pushed down to the client - when i did some training - it was 2gb.
this technology has to pre-download the model - then the streaming is fine....

thanks for any insights
if there's any open source projects that could help - i'm all ears.
i did see this for webcodecs
https://github.com/w3c/webcodecs/tree/main/samples

johndpope/IMF#38

UPDATE - I did some digging and found this NVC repo
microsoft/DCVC#63

did you use an onnx deployed to browser ? or just local python and consume streaming locally and run tests?