Timezone: »
DiffWave is 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. DiffWave produces high-fidelity audios in different waveform generation tasks, including neural vocoding conditioned on mel spectrogram, class-conditional generation, and unconditional generation. 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.
Author Information
Zhifeng Kong (University of California San Diego)
More from the Same Authors
-
2021 : Understanding Instance-based Interpretability of Variational Auto-Encoders »
· Zhifeng Kong · Kamalika Chaudhuri -
2021 : On Fast Sampling of Diffusion Probabilistic Models »
Zhifeng Kong -
2021 : Universal Approximation of Residual Flows in Maximum Mean Discrepancy »
Zhifeng Kong -
2022 : Panel Discussion »
Mirco Ravanelli · Chris Donahue · Zhifeng Kong · Wei-Ning Hsu · Rachel Manzelli · Sadie Allen