ChatTTS
ChatTTS is a generative text-to-speech model designed specifically for dialogue, such as the voice of an LLM assistant. Rather than reading text in a flat narration style, it aims for natural conversational speech: it supports multiple speakers and can predict and control fine-grained prosody, including laughter, pauses and interjections. It currently supports English and Chinese.
It is a good fit for researchers and developers prototyping voice assistants, conversational agents and dialogue demos, especially in Chinese and English. The main model was trained on more than 100,000 hours of audio; the open-source release on Hugging Face is a 40,000-hour pre-trained model without supervised fine-tuning. Streaming generation and zero-shot inference code are available, and the community maintains a list of end-user products built on it.
Getting started: install with `pip install ChatTTS`, or clone the repository and install its requirements (Python 3.11 is used in the conda example), then try the web UI with `python examples/web/webui.py`.
Features
Why choose it
Trade-offs
Maintenance check
- Last commit
- 2026-04-10
- Commits, last 90 days
- 0
- Releases, last 90 days
- 0
- Latest release
- v0.2.5 (2026-04-10)
- Open issues
- 62
- Recent issues with at least one comment
- 24/30
Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.
Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.
Most discussed open issues
- #640 使用vllm后的音色疑似只有男声15 comments · 2024-07-28
- #302 cost too many time!!!12 comments · 2024-06-12
- #284 How Can I use my own voice to training the AI8 comments · 2024-06-07
- #800 bug:main、dev code run main.py has error:lzma.LZMAError: Corrupt input data5 comments · 2024-10-24
- #503 生成的语音不全6 comments · 2024-07-01
Issue titles are quoted from GitHub. Open the link to check whether a problem still applies to the version you use.
Who it fits
- You need natural-sounding, conversational conversational text-to-speech synthesis with spontaneous pauses, sighs, and laughter.
- You want a self-hosted, open-weights TTS model optimized for conversational agents, podcasts, and dialog assistants in Python.
- You require commercial audio production with permissive licensing; the model weights are restricted under CC BY-NC 4.0 for non-commercial use.
- You need ultra-low-latency real-time voice-to-voice streaming with sub-200ms TTFT on minimal hardware.
How it differs
ChatTTS is a specialized conversational neural speech synthesis model. OpenChatCut is a full video editor combining multi-track timelines, media rendering, and conversational agent controls.
Compatibility
Quick start
Use cases
Alternatives
Comments
No comments yet. Be the first!