Diff-TTS: A Denoising Diffusion Model for Text-to-Speech
This work addresses the need for more natural and efficient TTS systems, which is incremental as it applies diffusion models to a known domain.
The authors tackled the problem of improving naturalness and efficiency in neural text-to-speech (TTS) by proposing Diff-TTS, a non-autoregressive model using a denoising diffusion framework, which achieved highly natural speech synthesis and generated 28 times faster than real-time on a single GPU.
Although neural text-to-speech (TTS) models have attracted a lot of attention and succeeded in generating human-like speech, there is still room for improvements to its naturalness and architectural efficiency. In this work, we propose a novel non-autoregressive TTS model, namely Diff-TTS, which achieves highly natural and efficient speech synthesis. Given the text, Diff-TTS exploits a denoising diffusion framework to transform the noise signal into a mel-spectrogram via diffusion time steps. In order to learn the mel-spectrogram distribution conditioned on the text, we present a likelihood-based optimization method for TTS. Furthermore, to boost up the inference speed, we leverage the accelerated sampling method that allows Diff-TTS to generate raw waveforms much faster without significantly degrading perceptual quality. Through experiments, we verified that Diff-TTS generates 28 times faster than the real-time with a single NVIDIA 2080Ti GPU.