F5-TTS
Open-source text-to-speech that clones a voice from a short sample clip and speaks new text in it.
- Stars
- 15.5k
- Forks
- 2.3k
- Licence
- MIT
- Language
- Python
- Last update
- 13 days ago
- Type
- AI model / agent
What it is
F5-TTS is an open-source model that turns written text into speech. You give it a short recording of a voice and the text that recording says, and it generates new speech in that voice saying whatever text you provide. It is official code from a research paper, and it comes with a web interface and a command-line tool so you can try it without writing much code.
- Generate speech from text in a cloned voice using the Gradio web interface
- Batch-generate speech from a text file using a TOML config with the CLI
- Finetune the model on your own voice data using the included training guidance
- Run it in Docker to serve a web interface on port 7860
Great for
- Researchers or developers who want to experiment with open voice-cloning text-to-speech
- People who want to generate narration in a voice they have permission to use
- Developers who want to self-host TTS rather than send text to a paid cloud service
- Anyone comfortable running Python and a GPU-backed setup
Not ideal for
- People who just want a finished voice app with no setup; the README assumes Python, pip, and a PyTorch install
- Users without a GPU who need fast results, since the README's benchmarks are all on a GPU
- Anyone wanting to clone voices of people who have not consented, as the README does not discuss consent or misuse policies
How to get it
- 1
Install with a package manager
You'll need: Python 3.10 or newer (README uses 3.11 in its example) · conda or virtualenv · FFmpeg · PyTorch with a build matched to your GPU
Create and activate a conda environment
conda create -n f5-tts python=3.11Install FFmpeg
conda install ffmpegInstall F5-TTS as a pip package (inference only)
pip install f5-tts - 2
Run it
- Install PyTorch with the build matching your device, as shown in the README
- Install the f5-tts pip package or clone the repo for editable install
- Launch the web interface with f5-tts_infer-gradio
- Open the local address in your browser and upload a reference voice clip with its transcript
- 3
Get the most out of it
- Use the --ref_text flag left empty to have the built-in ASR model transcribe the reference audio, at the cost of extra GPU memory
- Use a .toml config file with f5-tts_infer-cli -c to set up repeatable or multi-voice generation
- Read the inference guidance in src/f5_tts/infer to tune performance
- Use the Docker image with --gpus=all for a containerized GPU setup
Downloads
Strengths
- Open source under the MIT license with official code tied to a published paper
- Offers a web interface, a CLI, a pip package, and a Docker image
- Supports multiple models and a range of PyTorch setups including NVIDIA, AMD, Intel, and Apple Silicon
Trade-offs
- Requires a fairly technical setup, including Python 3.10+ and a matching PyTorch build for your GPU
- The README does not clearly describe ethical use limits or safeguards against misuse of voice cloning
- Model and dataset licensing for the pretrained weights is not summarized in the README excerpt
How it compares
| Project | Stars | Type | Level | Updated | Best for |
|---|---|---|---|---|---|
| F5-TTS (this) | 15.5k | AI model / agent | intermediate | 13 days ago | Researchers or developers who want to experiment with open voice-cloning text-to-speech |
| piper | 11.3k | App | intermediate | 1.2 years ago | Developers who want offline speech output in a local app or home automation setup |
| bark | 39.3k | AI model / agent | intermediate | 2.2 years ago | Developers who want to add text-to-speech or sound generation to a Python project |
Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.