#77th of 34 in Open modelsFor voice, speech and music
espnet
PyTorch toolkit for speech recognition, synthesis, translation and enhancement
Live demo ↗
(opens in a new tab)Documentation ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 10k
- License
- Apache-2.0
- Last commit
- Oct 2026
- Last release
- Sep 2026
- Language
- Python
9,738 on 31 Mar 20269,993 on 20 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
ESPnet is a PyTorch speech toolkit covering ASR, TTS, speech translation, enhancement, diarization, singing synthesis and speech language models. It provides Kaldi-style run.sh recipes for 200+ corpora and hundreds of pretrained models on Hugging Face. Inference is available through a CLI (espnet transcribe, synthesize, enhance), a Python API and an MCP server.
Who it is for: Speech researchers and engineers training or running ASR and TTS models
Strengths
- One recipe structure and trainer across ASR, TTS, ST, enhancement, SLU and more
- 200+ corpus recipes with run.sh from data prep to scoring
- Hundreds of pretrained models on Hugging Face, including OWSM v4 multilingual ASR
- pip install espnet gives CLI, Python and MCP inference without the training stack
Weaknesses
- Full recipe and training setup needs Kaldi-style tooling beyond pip install
- Recipe workflow assumes research familiarity with stages, corpora and configs
- ESPnet1 is unsupported; ESPnet2 and ESPnet3 recipes coexist
- Hardware requirements for training are not stated in the README
What it needs
- GPU optional
- Docker
- Needs PyTorch, Kaldi-style tooling (full recipe setup)
- Models: OWSM, Whisper, Conformer, E-Branchformer, VITS
Also in Open models
See all 34| Rank | Project | Score |
|---|---|---|
| 1 | llmfitRust tool that checks which LLMs fit your RAM, CPU and GPU | 87 out of 100 |
| 2 | layaSingle-pass multilingual classifier for typed choice, score and yes/no decisions | 84 out of 100 |
| 3 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 75 out of 100 |
| #4 | speechPyTorch toolkit for training and deploying ASR, TTS and speech LLM models | 70 out of 100 |
| #5 | yueLyrics-to-song model that writes an editable melody and chord score first | 65 out of 100 |