Add speech annotation tool using SAM-Audio

[AnkushMalaker]

[linear]

CHR-63

[DebjyotiRay]

1. Is this for automatic audio segmentation (using Meta's SAM-Audio to isolate speakers/remove noise), OR a manual annotation UI tool for labeling audio segments?
2. Should it be a new backend service, part of the existing advanced backend, or a web UI feature in the dashboard?

[AnkushMalaker]

Hey @DebjyotiRay
Sorry, forgot to update here- we've actually been discussing this a lot on discord.
To me, this changes the paradigm a bit of whats possible. Needs tweaking, but sure. I first want to know what this model is capable of, and for that, @01PrathamS tried to make a google colab notebook but wasn't able to load the model

I was able to load it on a 4090 and run some inference and the demo looked good. But its so big that its hard to run any experiments. I think its only a couple days though.
You should come ideate with us

Anyway-

1. Ideally one tool that does both seamlessly, that'd be the point.
2. This should be in extras/ as a separate micro service with its own UI. The service can be exposed to the backend via some API. The speaker recognition service is arranged that way.

But still, try running the model on something and let me know if you're able to

[DebjyotiRay]

Would love to ideate, where's the discord? Let's see if we can come to someplace in between towards the implementation

[01PrathamS]

Chronicle AI Discord: https://discord.com/channels/1389196827064008755/1445389493568929865

join in here @DebjyotiRay