Skip to content

Reverb

Open-source ASR and speaker diarization code

Visit official site

What Reverb is

The project is described as open-source inference code for speech recognition and speaker diarization on long-form audio.

What you can do with it

  • Transcribe a meeting and split speakers
  • Analyse call centre recordings
  • Transcribe and identify speakers in court records

What to watch out for

  • Git LFS must be installed first
  • A HUGGINGFACE_ACCESS_TOKEN is needed for model download
  • A virtual environment is set up by the user

Pros & cons

✓ What we like

  • Models can be downloaded from Hugging Face Hub
  • NVIDIA GPU support is listed

! What to watch out for

  • Git LFS must be installed first
  • A HUGGINGFACE_ACCESS_TOKEN is needed for model download

FAQ

Which frameworks are named?

WeNet is used for ASR and Pyannote for speaker diarization.

How are models obtained?

Models are downloaded from Hugging Face Hub with a token.

Last reviewed: 2026-11-15

More AI audio tools tools

View all →

How we review