Remote Whisper model

[utrack]

Hi there!

There is a Faster Whisper Server project (https://github.com/fedirz/faster-whisper-server) which wraps the local Whisper model and opens it up to the local network.

Basically, my laptop's GPU is too bad to run the model on it, so I'd like to hook local WhisperX with some initial audio processing (VAC, DAC, silence filtering) into a remote model (the actual voice recog & diarization).

Before I start hacking - maybe someone already tried that?

Would it make sense to try it to see improvements compared to https://github.com/ufal/whisper_streaming ?

[fraschm1998]

someone made: https://github.com/namastexlabs/whisperx-api

but I haven't tried it yet.