avion23

Thanks that seems to work on TWRP. hardware is mi 10 pro, software recovery-TWRP-3.4.2B-0623-XIAOMI10PRO-CN-wzsx150

pyannote/speaker-diarization-3.1.1 has more parameters, but it turns out they are read only :( Probably that's only documentation on why the model was trained.

I'll check again this weekend and upload photos