Open-source text-to-speech
F5-TTS
Flow-matching speech synthesis that uses a reference recording and its transcript for zero-shot voice cloning.
- Zero-shot cloning
- Reference transcript
- Flow matching
Version & availability
- Project release
- v1 Base
- Embedded demo
- F5-TTS
- Languages
- Chinese and English with the base checkpoints
Upstream's current base model is F5-TTS v1 Base. The community demo does not explicitly identify its deployed checkpoint version.
Last checked:
Try F5-TTS
Demo: F5-TTS · Community demo linked by the project
Hosted on Hugging Face. GPU queues, sign-in or usage limits may apply. If the demo does not load here, open it directly using the link above.
License & model terms
Code: MIT. The project's pretrained models are licensed CC-BY-NC due to the training data.
Read the project documentation ↗