A Text To Speech model that can clone sounds. The CPU runs slowly, so it is suitable for CPU performance testing. ##Characteristics 1. ** Zero-sample & small-sample TTS**: High-quality TTS output can be generated by inputting sound samples for 10 to 30 seconds. ** For details, see [Speech Cloning Best Practices Guide](https://docs.fish.audio/text-to-speech/voice-clone-best-practices). ** 2. ** Multi-language and cross-language support **: Just copy and paste multi-language text into the input box, no need to worry about language issues. Currently supports English, Japanese, Korean, Chinese, French, German, Arabic and Spanish. 3. ** No phoneme dependence **: The model has strong generalization capabilities, does not rely on phonemes for TTS, and can process any language represented by text. 4. ** High accuracy rate **: A CER (character error rate) and WER (word error rate) of approximately 2% were achieved on 5 minutes of English text. 5. ** Fast **: With fish-tech acceleration, the real-time factor is approximately 1:5 on Nvidia RTX 4060 notebooks and approximately 1:15 on Nvidia RTX 4090. 6. **WebUI Inference **: Provides an easy-to-use Gradio-based web user interface, compatible with Chrome, Firefox, Edge and other browsers. 7. **GUI reasoning **: Provides PyQt6 graphical interface and seamlessly cooperates with API servers. Supports Linux, Windows and macOS. [Check the GUI](https://github.com/AnyaCoder/fish-speech-gui). 8. ** Easy to deploy **: Easily set up an inference server, natively supports Linux, Windows and macOS, minimizing speed losses. ##Disclaimer We accept no responsibility for any illegal use of the code base. Please refer to your local information on DMCA (Digital Millennium Act) and other relevant laws and regulations.
