WaveNet: A Generative Model for Raw Audio

WaveNet: A Generative Model for Raw Audio is a 2016 paper from DeepMind by Aäron van den Oord and others, which introduced Wavenet.

Notes from paper WaveNet: A Generative Model for Raw Audio

WaveNet generates raw audio waveforms one sample at a time, with each sample conditioned on all the previous samples (autoregressive). To cover a long enough history of samples, it uses stacks of dilated causal convolutions, whose receptive field grows exponentially with depth.

It produced much more natural-sounding text-to-speech than previous systems, and can also generate music. It was later used as the vocoder in Natural TTS Synthesis by Conditioning WaveNet on Mel Spectrogram Predictions.