F5-TTS
SWividA flow-matching diffusion-transformer text-to-speech model for zero-shot voice cloning in English and Chinese, released with its training and fine-tuning code. The project is led by SJTU X-LANCE researchers under the SWivid account.
The code is MIT; the pretrained weights are CC-BY-NC-4.0 because they were trained on the Emilia dataset.
Openness
2 high confidence- weights
- open(checkpoints on the Hub, ungated)
- data
- open(trained on the public Emilia dataset)
- code
- open(training and fine-tuning code in the repository)
- license
- CC-BY-NC-4.0(weights
F5-TTS publishes its code under MIT and trains on the public Emilia corpus, but the weights carry a non-commercial license inherited from that data. Downloadable is not usable commercially, and that is what sets the score.
- https://huggingface.co/api/models/SWivid/F5-TTS recorded 2026-09-27
"license:cc-by-nc-4.0" tag and "gated":false.
- https://huggingface.co/SWivid/F5-TTS/raw/main/README.md recorded 2026-09-27
Card front matter "license: cc-by-nc-4.0" and "datasets:" naming "amphion/Emilia-Dataset".
- https://raw.githubusercontent.com/SWivid/F5-TTS/HEAD/README.md recorded 2026-09-27
"Our code is released under MIT License. The pre-trained models are licensed under the CC-BY-NC license due to the training data Emilia"; "Refer to [training & finetuning guidance](src/f5_tts/train) for best practice."
Adoption
3 high confidenceHugging Face downloads of the weights repository. The f5-tts package is marked as not a separate channel because it loads this repository.
- https://huggingface.co/api/models/SWivid/F5-TTS recorded 2026-09-27
"downloads":905144 for SWivid/F5-TTS.
Capability
3 low confidenceAn influential open voice-cloning architecture, limited to two languages. It is not on the Artificial Analysis TTS leaderboard, so it is placed on its own reported results on a shared public test set, level with Kokoro, rather than on a blind listening test.
- https://arxiv.org/pdf/2410.06885 recorded 2026-09-27
Paper (PDF; read with a text extractor): Table 6 gives evaluation results of F5-TTS on LibriSpeech-PC test-clean, Seed-TTS test-en and Seed-TTS test-zh, and Table 2 compares it with other systems on the two Seed-TTS test sets.
Verified 2026-09-27