Non-Attentive Tacotron: Robust and Controllable Neural TTS Synthesis Including Unsupervised Duration Modeling
This work addresses robustness and control issues in TTS for applications requiring reliable synthesis, though it is incremental as it builds on Tacotron 2.
The paper tackles the problem of improving robustness and controllability in neural text-to-speech synthesis by replacing the attention mechanism in Tacotron 2 with an explicit duration predictor, resulting in enhanced metrics like a 4.41 mean opinion score for naturalness and reduced word deletion rates.
This paper presents Non-Attentive Tacotron based on the Tacotron 2 text-to-speech model, replacing the attention mechanism with an explicit duration predictor. This improves robustness significantly as measured by unaligned duration ratio and word deletion rate, two metrics introduced in this paper for large-scale robustness evaluation using a pre-trained speech recognition model. With the use of Gaussian upsampling, Non-Attentive Tacotron achieves a 5-scale mean opinion score for naturalness of 4.41, slightly outperforming Tacotron 2. The duration predictor enables both utterance-wide and per-phoneme control of duration at inference time. When accurate target durations are scarce or unavailable in the training data, we propose a method using a fine-grained variational auto-encoder to train the duration predictor in a semi-supervised or unsupervised manner, with results almost as good as supervised training.