Parallel Tacotron 2: A Non-Autoregressive Neural TTS Model with Differentiable Duration Modeling
This work addresses text-to-speech synthesis for applications needing efficient and controllable speech generation, though it is incremental as it builds on prior non-autoregressive models.
The paper tackles the problem of generating speech from text without requiring supervised duration signals, by introducing Parallel Tacotron 2, a non-autoregressive neural TTS model with a differentiable duration model; it shows improved subjective naturalness over baselines in multi-speaker evaluations.
This paper introduces Parallel Tacotron 2, a non-autoregressive neural text-to-speech model with a fully differentiable duration model which does not require supervised duration signals. The duration model is based on a novel attention mechanism and an iterative reconstruction loss based on Soft Dynamic Time Warping, this model can learn token-frame alignments as well as token durations automatically. Experimental results show that Parallel Tacotron 2 outperforms baselines in subjective naturalness in several diverse multi speaker evaluations. Its duration control capability is also demonstrated.