F5-TTS
SWivid/F5-TTSReviewed by AI
Checked by RepoPicks Radar on
What is F5-TTS?
F5-TTS is the research code of a text-to-speech model that copies a voice. You give it a short reference recording with its transcript, type the text you want, and it produces audio in that voice, in English or Chinese and other languages added by the community. It opens a web page on your own machine (Gradio) or runs from the terminal, and it can also be trained on new voices. You install it with Python and PyTorch, and a graphics card is strongly recommended.
F5-TTS is free and open source (MIT) and can replace ElevenLabs, Respeecher, and Murf.
How easy is it to install F5-TTS?
D · terminal, several tools or a specific graphics card · Windows
- Windows D
- macOS D
- Linux D
On all three systems you create a Python 3.10+ environment, install FFmpeg, pick the PyTorch build that matches your hardware (CUDA, ROCm, Intel XPU or Apple Silicon) and then "pip install f5-tts" before typing "f5-tts_infer-gradio" to open the local web page. There is also a Docker image, which still needs an NVIDIA card with --gpus=all.
Coming soon: install in 1 click with the RepoPicks app.
F5-TTS on GitHub
- Stars
- 15,560on Nov 10, 2026
- Downloads
- —all releases
- Growth
- +48stars in 6 days
Stars over time
Facts
- Runs on
- Windows, macOS, Linux
- You need to install
- Python, FFmpeg, Graphics card
- To build from source
- Python, Docker, Graphics card, FFmpeg
- Disk space
- 950 MB–4.3 GB · estimate
- Accounts or keys
- None
- Cost
- Free
- License
- MIT (Permissive)
- Can replace
- ElevenLabs, Respeecher, Murf, Kits AI, Fish Audio
- Last change on GitHub
- Nov 3, 2026
F5-TTS risk signals (5)
- Attentionmodels not free for business
README.md License: "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, which is an in-the-wild dataset." The voices you generate come from models that forbid commercial use.
AI review
- Attentionclones anyone s voice
The CLI takes "--ref_audio 'provide_prompt_wav_path_here.wav'" plus its transcript and speaks any text in that voice, and the project title is "A Fairytaler that Fakes Fluent and Faithful Speech". Nothing in the repository checks consent, and the watermark image in the README is commented out.
AI review
- Infodownloads models on first use
The dependency "cached_path" and the Docker volume "/root/.cache/huggingface/hub/" show the model files are fetched from Hugging Face the first time you run it; the Gradio app also offers "Voice Chat powered by Qwen2.5-3B-Instruct", another large download.
AI review
- Infopublic link option
README.md shows "f5-tts_infer-gradio --share" to "Launch a share link", which exposes the interface running on your machine through a public address.
AI review
- Inforesearch code
The repository is the "Official code" for the arXiv paper 2410.06885, with training, evaluation and fine-tuning folders. It is published for researchers, not packaged as a finished app.
AI review
Evidence is shown in its original language.
What Radar could not check about F5-TTS
- How close the cloned voice sounds to the original recording
- Whether it runs at a usable speed without an NVIDIA card, for example on a Mac
- The total disk space after the model files are downloaded
- Which languages beyond English and Chinese work well, since they come from community models
- How much memory is needed when the reference text is transcribed automatically