gemelo-ai/vocos
A neural vocoder that synthesizes high-quality audio waveforms from mel-spectrograms or EnCodec tokens using a GAN-based approach.
Not currently ranked — collecting fresh signals.
star history
Vocos is a fast neural vocoder that generates audio waveforms from acoustic features in a single forward pass. Unlike typical time-domain GAN vocoders, it generates spectral coefficients which are rapidly converted to audio via inverse Fourier transform. It supports inference from mel-spectrograms and EnCodec quantization tokens, with pretrained models available at 24kHz.
Frequently asked
- What is gemelo-ai/vocos?
- A neural vocoder that synthesizes high-quality audio waveforms from mel-spectrograms or EnCodec tokens using a GAN-based approach.
- Is vocos open source?
- Yes — gemelo-ai/vocos is open source, released under the MIT license.
- What language is vocos written in?
- gemelo-ai/vocos is primarily written in Python.
- How popular is vocos?
- gemelo-ai/vocos has 1.1k stars on GitHub.
- Where can I find vocos?
- gemelo-ai/vocos is on GitHub at https://github.com/gemelo-ai/vocos.