TL;DR Better pretrained pipeline and model Much better overlapping speech detection with powerset pyannote/segmentation-3.0 Much better speaker diarization performance with pyannote/speaker-diariz...