Simple and Effective Multi-sentence TTS with Expressive and Coherent Prosody
This addresses the problem of poor prosody in long-form TTS for users needing natural-sounding speech, though it is incremental as it builds on existing methods.
The paper tackled the challenge of generating expressive and coherent prosody for multi-sentence text-to-speech by extending a Transformer-based system, achieving statistically significant improvements in speech naturalness over competitors.
Generating expressive and contextually appropriate prosody remains a challenge for modern text-to-speech (TTS) systems. This is particularly evident for long, multi-sentence inputs. In this paper, we examine simple extensions to a Transformer-based FastSpeech-like system, with the goal of improving prosody for multi-sentence TTS. We find that long context, powerful text features, and training on multi-speaker data all improve prosody. More interestingly, they result in synergies. Long context disambiguates prosody, improves coherence, and plays to the strengths of Transformers. Fine-tuning word-level features from a powerful language model, such as BERT, appears to profit from more training data, readily available in a multi-speaker setting. We look into objective metrics on pausing and pacing and perform thorough subjective evaluations for speech naturalness. Our main system, which incorporates all the extensions, achieves consistently strong results, including statistically significant improvements in speech naturalness over all its competitors.