Morgan Funtowicz

Thanks @vectorventures ! Word-level timestamps is not yet supported and from my readings, even through OAI platform with Whisper it's not perfect either.

I was looking at the CrisperWhisper @razhan mentionned and it seems to provide some good/better word-level timestamps out of the box, so maybe this is something we can investigate in the coming weeks? would you like to take a look and potentially open a PR for this?

@RavirajDarisi Of course, you can deploy your own endpoint for instance, and use the URL within your Android app 👌- lmk if you need some more guidance.

We do provide a precompiled image if you want: docker pull mfuntowicz/endpoints-whisper-vllm:v1.0.2-py312

To run with Docker you should invoke with something like this:

$> docker run --gpus 1 -p 8000:80 -e MODEL_ID=openai/whisper-large-v3-turbo mfuntowicz/endpoints-whisper-vllm:v1.0.2-py312

It will starts the container exposing port 8000 on your local machine. You can for instance navigate to http://localhost:8000/docs to see documentation.

If you want some more information, the repository @Younggun linked should provide some more and of course, feel free to reply here 🤗

If you pull the Docker image locally and have a GPU available on your computer/server, it should yes 👌

I can give it a try in the coming days!