#6565th of 88 in VoiceFor voice, speech and music
wenet
Speech recognition toolkit for training and deploying ASR models
- Stars
- 5.3k
- License
- Apache-2.0
- Last commit
- Oct 2026
- Last release
- May 2024
- Language
- Python
5,069 on 21 Apr 20265,256 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
WeNet is a PyTorch toolkit for end-to-end speech recognition, with training recipes, a pip package for transcription and a C++ runtime for deployment. The pip package offers a CLI and Python API with Paraformer, FireRed and WeNetSpeech models for Chinese and Whisper large-v3 variants for English. Language model integration uses WFST decoding, which needs the runtime built from source.
Who it is for: Engineers training or deploying speech recognition, especially Chinese ASR
Strengths
- Covers training, Python inference and a separate C++ runtime in one repository
- Supports both streaming and non-streaming recognition models
- Huawei Ascend NPU install path documented alongside CUDA
- Pretrained models and a HuggingFace demo are linked
Weaknesses
- CLI runs on CPU by default; GPU needs an explicit --device cuda flag
- Unpinned torch dependency can install a build that mismatches the GPU driver
- Last tagged release is 2024-05-23; pip install pulls from git
- Runtime and LM decoding need a manual cmake build; no Docker image
What it needs
- GPU optional
- Needs PyTorch, torchaudio, sox, conda, CUDA 12.1 (recommended for training), cmake 3.14+ (runtime build)
- Models: paraformer, firered, wenetspeech, whisper-large-v3, whisper-large-v3-turbo
What changed
From the release notes. No flag means the note does not say so, not that the upgrade is safe.
Also in Voice
See all 88| Rank | Project | Score |
|---|---|---|
| 1 | voicestudioDesktop app for voice cloning, dubbing, dictation and audiobooks | 98 out of 100 |
| 2 | handyOffline push-to-talk dictation app that pastes transcripts into any text field | 91 out of 100 |
| 3 | whisper.cppC/C++ speech-to-text engine running OpenAI Whisper models locally | 90 out of 100 |
| #4 | buzzOffline desktop app for transcribing and translating audio with Whisper | 84 out of 100 |
| #5 | xiaozhi-esp32ESP32 firmware for a voice chatbot with MCP device and cloud tool control | 83 out of 100 |