In simple English
This tool takes written text and turns it into spoken audio files. It speaks over six hundred different languages with natural intonation. You can give it a short recording of any person and it will copy their exact voice. It runs on your own computer or server using a graphics card. It is free to download and use under an open source licence.
What it is
OmniVoice is a massively multilingual text-to-speech model built on a diffusion language model architecture. It provides Python libraries and command-line tools for zero-shot voice cloning, voice design, and batch audio generation.
What it is good for
Solo builders who need to generate spoken audio in multiple languages or clone specific voices for podcasts, videos, or applications without paying commercial cloud fees.
- Supports over six hundred languages with high fidelity.
- Performs zero-shot voice cloning from a short three to ten second reference audio clip.
- Includes command-line tools and a local web interface for quick testing.
- Requires a dedicated graphics card with several gigabytes of memory to run locally.
- Voice design features are primarily trained on Chinese and English data.
- Large model downloads and PyTorch dependency installation take time and disk space.
Commercial cloud text-to-speech APIs
Hugging Face transformers
- myshell-ai/OpenVoice38k stars
Another open source voice cloning model focused on instant tone color conversion.
- fishaudio/fish-speech33k stars
An open source text-to-speech model supporting multilingual generation and voice cloning.
- shaka-project/shaka-packager2.6k stars
Cuts a video file into the pieces and index files that a browser needs in order to stream it.
- mpv-player/mpv38k stars
A command-line media player that handles most audio, video and subtitle formats and can be driven entirely from scripts.
- snapcast/snapcast8.0k stars
A server and clients that play one audio stream in several rooms, kept in time with each other.
git clone https://github.com/k2-fsa/OmniVoice.git cd OmniVoice pip install -e .