# wenet-e2e/wespeaker

Research and Production Oriented Speaker Verification, Recognition and Diarization Toolkit

Repository: https://github.com/wenet-e2e/wespeaker
Canonical: https://ross.abutalabs.com/products/wespeaker
Language: Python
License: Apache-2.0
License Family: permissive
Topics: asv, ecapa-tdnn, production-ready, pytorch, resnet, speaker-recognition, speaker-verification, xvector, speaker-diarization, repvgg, campplus, eres2net, ssl, dino, plda, cnceleb, voxceleb, nist-sre, wavlm, redimnet
Last push: 2026-07-08T04:25:34+00:00

## Health v2 (maintenance only)
Score: 64/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 91, release rhythm 8, longevity 100
- inputs: {"age_days": 1800, "days_push": 56, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1392, forks 201 (observed 2026-08-28T04:04:36.003234+00:00)

## What it is
WeSpeaker is a research and production-oriented toolkit for speaker embedding learning, supporting speaker verification, recognition, and diarization. It provides pretrained models (xvector, ECAPA-TDNN, ResNet, CAM++, and more), a Python API, and command-line tools built on PyTorch.

## Use cases
- extract speaker embeddings from audio files
- verify whether two voice recordings belong to the same speaker
- perform speaker diarization on a single audio file
- train custom speaker verification models on VoxCeleb or CN-Celeb
- deploy speaker recognition models to production with runtime support
- compare speaker similarity between two wav files
- batch extract embeddings for a list of utterances

## When to choose
- you need state-of-the-art speaker verification or diarization with pretrained models
- you want both research training pipelines and production deployment runtimes
- you prefer a simple Python API or CLI over building models from scratch

## When to avoid
- you need general speech recognition (ASR) or text transcription rather than speaker identity
- you work outside Python/PyTorch ecosystems and need a pure C++ or mobile-first solution
- you only need lightweight voice activity detection without speaker modeling

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, audio-processing, speech-recognition, cli, sdk
- domain: speech-processing, machine-learning, developer-tools
- platform: python, cli
- tags: speaker-verification, speaker-embedding, speaker-diarization, pytorch, xvector, ecapa-tdnn, resnet, voxceleb, pretrained-models, self-supervised-learning, audio, linux, macos, gpu

## Member repositories
- wenet-e2e/wespeaker (main) score 64

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:36.003234+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 2026-08-30T04:39:34.375351+00:00, confidence not recorded.
  - readme: https://github.com/wenet-e2e/wespeaker (fetched 2026-08-28T04:04:36.003234+00:00, sha 9d0566c4bedb)
- Data as of 2026-08-30T08:39:29.467469+00:00.