Kokoro
A tiny 82M-parameter TTS model with surprisingly natural voices.
- Star 数
- 9.3K
- Fork 数
- 1K
- 未关闭 Issue
- 214
- 最近提交
- 2025-08-06
数据来自 GitHub API · 更新于 2026-10-15
Kokoro 是什么？
Kokoro is a lightweight open-weight TTS model that runs fast even on CPU while producing natural speech. Apache-licensed weights make it a popular choice for apps and local assistants.
主要功能
- Only 82M parameters — runs on CPU
- Multiple voices and languages
- Apache-2.0 weights
系统要求
- 平台
- Windows · macOS · Linux
- Python
- 3.10+
- 备注
- Requires espeak-ng for some languages.
如何安装 Kokoro
根据项目 README 整理。开源项目更新很快，请以 GitHub 页面上的最新说明为准。
pip
- 1Install the package (and espeak-ng via your package manager).
pip install kokoro soundfile - 2Generate speech in Python.
from kokoro import KPipeline import soundfile as sf pipeline = KPipeline(lang_code="a") for i, (_, _, audio) in enumerate(pipeline("Hello from Kokoro!", voice="af_heart")): sf.write(f"out_{i}.wav", audio, 24000)
优势
- Fast and small
- Commercially friendly license
不足
- No voice cloning
Kokoro 替代品对比
| 软件 | Star 数 | 许可证 | 平台 | 显存 |
|---|---|---|---|---|
| Kokoro A tiny 82M-parameter TTS model with surprisingly natural voices. | 9.3K | Apache-2.0 | — | |
| F5-TTS Zero-shot voice cloning TTS based on flow matching. | 15.4K | MIT | — | |
| Coqui TTS Deep-learning text-to-speech toolkit, including the XTTS voice cloning model. | 46.1K | MPL-2.0 | — | |
| ChatTTS Conversational TTS for dialogue, with laughter and pauses. | 39.9K | AGPL-3.0 | 4 GB+ | |
| Whisper OpenAI’s open-source speech recognition and translation model. | 110.5K | MIT | ~1 GB (tiny) to ~10 GB (large) | |
| whisper.cpp Whisper in plain C/C++ — fast on CPUs, Apple Silicon and mobile. | 54.4K | MIT | — |
常见问题
Kokoro 免费吗？
是的——Kokoro 是开源软件，采用 Apache-2.0 许可证。商用前请确认软件及模型权重的许可条款。
Kokoro 支持 Windows、Mac 和 Linux 吗？
Kokoro 支持：Windows, macOS, Linux。
如何安装 Kokoro？
1. Install the package (and espeak-ng via your package manager). 2. Generate speech in Python.
Kokoro 有哪些好的替代品？
常见替代品包括 F5-TTS, Coqui TTS, ChatTTS, Whisper，详见本页对比表。