Can someone share their python environment that works with whisperx for speaker diarization? I keep getting errors even when using the examples from the github.

[GreatestJakeEVR]

So I'm trying to take an audio recording of my work meetings and use AI to create a transcription. I have near a decade of python experience and about 5 years of experience playing around with data science and ML.

I tried to get whisperx to work, but I always get an error. Sometimes its something to do with numpy changing NaN to nan, other times its an error with pyannote/Speaker-Diarization-3.0. Multiple times I've created completely new environments and only installed the packages that were essential to running whisper, but still I can't get it to work.

I followed the setup and usage section of the github readme and the commands would always fail. I also tried using the "Python Usage" example and that also failed.

So if anyone could share with me their python environment (the python version and the packages you have installed) and a list of software along with their versions that I'd need installed on my PC (for example FFmpeg) that would really help me out a lot.

I do my coding in WSL2 with an Ubuntu installation and use miniconda for my venvs if that matters any. If someone has a working way to use whisperx that would require some other way of doing it that is fine also. I've used just about every possible way of installing python and managing packages that there is lol.

[adamlogan]

I would love this too. Whisper was working well for me on Python 3.10.

If you can't get WhisperX working well on your machine, you can always try a docker container.