Reverb provides open source inference code for Rev’s ASR and diarization models. The ASR component uses the WeNet framework and the speaker diarization component uses the Pyannote framework. Models are available on HuggingFace and benchmarks compare performance against Whisper and Canary on standard datasets.