Waveglow. Generative modeling for audio synthesis

Talk presentation

Waveglow - fast, parallel non-autoregressive flow-based generative neural network. Combines insights from Glow and WaveNet and trained using single loss-function which maximizes the likelihood of the training data.

Capable to produce up to 22 faster then real-time high quality audio samples.

Taras Sereda
Researcher
  • Machine learning researcher and entrepreneur, working on speech synthesis.
  • Enthusiastic about math theory, linear algebra, neural nets learning techniques.
  • Twitter
Sign in
Or by mail
Sign in
Or by mail
Register with email
Register with email
Forgot password?