Can I use this as drop in replacement for pyannote/speaker-diarization-3.1?
Hi,
Thanks for making this open source. I have a (beginner) question. So I run my transcription pipeline with: WhisperX
Which uses the pyannote/speaker-diarization-3.1... Will I be able to use your model as a replacement for that?
I already tried this. One thing I noticed is that this repo lacks the config.yaml file like the pyannote/speaker-diarization-3.1 has. So when I try to use it as a drop-in replacement, it's giving me a bug. Is it possible to add a config.yaml file to this perhaps? Or am I thinking about this the wrong way?
Maybe it should be the values here : https://github.com/revdotcom/reverb/blob/main/diarization/infer_pyannote3.0.py#L17
That should be written in this YAML format : https://huggingface.co/pyannote/speaker-diarization-3.1/resolve/main/config.yaml and committed to this repo?
If you can confirm this is the fix, I am happy to send a PR.
Hi,
I appreciate your interest in our diarization model.
I think you are right, we initialize the pipeline manually in the code, which is not the best practice and if you want to just replace the pyannote pipeline, it will not work.
I think that we would need to release another "model", which would be something like reverb-diarization-pipeline-v2. Let me discuss this with my colleagues, but I don't see a reason why not.
I am not sure if config.yaml file would work inside this specific model - it would point to itself.
Hi,
so I went ahead and added the config as you described - there was no need to create another model with pipeline, sorry for the confusion.
It should work now, lmk if you have issues.
There is also PR to fix github tools: https://github.com/revdotcom/reverb/pull/10
Looks awesome. Thanks a lot for fixing it up so quickly. I will check soonish and report back. Thank you again.