Model reference · open weights

reverb-diarization

reverb-diarization is an open-weight audio or speech model from Revai. reverb-diarization-v1 (BF16) weighs 28 MB; the smallest configuration that runs it is RTX 3060 12 GB.

• Reverb diarization V1 is an automatic speech recognition model developed by Revai for speaker diarization tasks.
• It achieves a 16.5% relative improvement in Word Diarization Error Rate compared to the pyannote3.0 baseline, evaluated on over 1,250,000 tokens across five test suites.
• The model is released under an "other" license.

Summary of the Revai/reverb-diarization-v1 model card, 2026-10-01

Hugging Face ↗Paper ↗