# pyannote/pyannote-audio

Neural building blocks for speaker diarization: speech activity detection, speaker change detection, overlapped speech detection, speaker embedding

Repository: https://github.com/pyannote/pyannote-audio
Canonical: https://ross.abutalabs.com/products/pyannote-audio
Homepage: https://www.pyannote.ai
Language: Jupyter Notebook
License: MIT
License Family: permissive
Topics: pytorch, speech-processing, speaker-diarization, speech-activity-detection, speaker-change-detection, speaker-embedding, voice-activity-detection, pretrained-models, overlapped-speech-detection, speaker-recognition, speaker-verification
Last push: 2026-08-04T17:30:45+00:00

## Health v2 (maintenance only)
Score: 95/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 96, release rhythm 90, longevity 100
- inputs: {"age_days": 3831, "days_push": 29, "days_rel": 64, "gap_med": 18.5, "n_releases_24m": 9}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 10475, forks 1101 (observed 2026-08-28T04:10:42.080323+00:00)

## What it is
pyannote.audio is an open-source Python toolkit built on PyTorch for speaker diarization, providing neural building blocks like voice activity detection, speaker change detection, overlapped speech detection, and speaker embeddings. It ships state-of-the-art pretrained models and pipelines via Hugging Face that can be finetuned on custom data.

## Use cases
- identify who spoke when in an audio recording
- detect voice activity in audio files
- detect overlapping speech segments
- extract speaker embeddings for voice recognition
- finetune speaker diarization models on my own data
- segment meeting transcripts by speaker

## When to choose
- you need local, self-hosted speaker diarization with pretrained models
- you want to finetune diarization components on your own audio data
- you need building blocks like VAD or speaker embeddings for a voice AI pipeline

## When to avoid
- you want a fully managed API without running models yourself (consider pyannoteAI premium)
- you need speech-to-text transcription rather than diarization
- you work outside the Python/PyTorch ecosystem

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, speech-recognition, audio-processing, deep-learning
- domain: speech-processing, machine-learning
- platform: python, cross-platform
- tags: speaker-diarization, voice-activity-detection, pytorch, pretrained-models, speaker-embedding, pyannote, audio, natural-language-processing, gpu

## Member repositories
- pyannote/pyannote-audio (main) score 95

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:42.080323+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:18:59.224736+00:00, confidence not recorded.
  - readme: https://github.com/pyannote/pyannote-audio (fetched 2026-08-28T04:10:42.080323+00:00, sha 1fe6d3ca91ff)
  - homepage: https://www.pyannote.ai (fetched 2026-08-29T08:18:08.794939+00:00, sha b4d8bb84999c)
  - site_page: https://docs.pyannote.ai/introduction (fetched 2026-08-29T08:18:08.806765+00:00, sha b726a04b9c4c)
  - site_page: https://www.pyannote.ai/pricing (fetched 2026-08-29T08:18:08.804446+00:00, sha fdaa642e5f0c)
- Data as of 2026-08-30T08:39:29.467469+00:00.