NU-Wave: A Diffusion Probabilistic Model for Neural Audio Upsampling
This work addresses audio quality enhancement for applications like speech synthesis and audio restoration, representing a novel advancement in high-resolution audio generation.
The paper tackles the problem of neural audio upsampling by introducing NU-Wave, a diffusion probabilistic model that generates 48kHz waveforms from 16kHz or 24kHz inputs, outperforming baselines in SNR, LSD, and ABX tests with a smaller model size of 3.0M parameters.
In this work, we introduce NU-Wave, the first neural audio upsampling model to produce waveforms of sampling rate 48kHz from coarse 16kHz or 24kHz inputs, while prior works could generate only up to 16kHz. NU-Wave is the first diffusion probabilistic model for audio super-resolution which is engineered based on neural vocoders. NU-Wave generates high-quality audio that achieves high performance in terms of signal-to-noise ratio (SNR), log-spectral distance (LSD), and accuracy of the ABX test. In all cases, NU-Wave outperforms the baseline models despite the substantially smaller model capacity (3.0M parameters) than baselines (5.4-21%). The audio samples of our model are available at https://mindslab-ai.github.io/nuwave, and the code will be made available soon.