shivammehta25/Matcha-TTS
A neural text-to-speech model that generates speech from text using conditional flow matching.
Not currently ranked — collecting fresh signals.
star history
Matcha-TTS is a non-autoregressive neural TTS system that uses conditional flow matching to synthesize speech from text. The model learns a probability path between noise and audio through diffusion-style training, and performs inference by solving an ODE to generate waveforms. The system is designed to be fast, probabilistic, and memory-efficient while producing natural-sounding speech, published at ICASSP 2024.
Frequently asked
- What is shivammehta25/Matcha-TTS?
- A neural text-to-speech model that generates speech from text using conditional flow matching.
- Is Matcha-TTS open source?
- Yes — shivammehta25/Matcha-TTS is open source, released under the MIT license.
- What language is Matcha-TTS written in?
- shivammehta25/Matcha-TTS is primarily written in Jupyter Notebook.
- How popular is Matcha-TTS?
- shivammehta25/Matcha-TTS has 1.3k stars on GitHub.
- Where can I find Matcha-TTS?
- shivammehta25/Matcha-TTS is on GitHub at https://github.com/shivammehta25/Matcha-TTS.