DiffWave: A Versatile Diffusion Model for Audio Synthesis

09/21/2020 ∙ by Zhifeng Kong, et al. ∙ 9

In this work, we propose DiffWave, a versatile Diffusion probabilistic model for conditional and unconditional Waveform generation. The model is non-autoregressive, and converts the white noise signal into structured waveform through a Markov chain with a constant number of steps at synthesis. It is efficiently trained by optimizing a variant of variational bound on the data likelihood. DiffWave produces high-fidelity audios in Different Waveform generation tasks, including neural vocoding conditioned on mel spectrogram, class-conditional generation, and unconditional generation. We demonstrate that DiffWave matches a strong WaveNet vocoder in terms of speech quality (MOS: 4.44 versus 4.43), while synthesizing orders of magnitude faster. In particular, it significantly outperforms autoregressive and GAN-based waveform models in the challenging unconditional generation task in terms of audio quality and sample diversity from various automatic and human evaluations.



There are no comments yet.


page 1

page 2

page 3

page 4

Code Repositories


DiffWave is a fast, high-quality neural vocoder and waveform synthesizer.

view repo


Pytorch Reimplementation of DiffWave Vocoder: a high quality, fast, and small neural vocoder.

view repo


Tensorflow implementation of DiffWave: A Versatile Diffusion Model for Audio Synthesis

view repo


DiffWave with variaitional diffusion models

view repo
This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.