VoiceStar: Robust Zero-Shot Autoregressive TTS with Duration Control and Extrapolation

TL;DR

VoiceStar introduces a zero-shot TTS model with advanced duration control and extrapolation capabilities, leveraging novel embeddings and training techniques to improve long-form speech synthesis quality.

Contribution

The paper presents VoiceStar, the first zero-shot TTS model capable of duration control and extrapolation, using PM-RoPE and CPM training for improved alignment and speech quality.

Findings

Outperforms state-of-the-art on long-form speech benchmarks

Achieves better speaker similarity and intelligibility in extrapolation tasks

Maintains competitive performance on short-form benchmarks

Abstract

We present VoiceStar, the first zero-shot TTS model that achieves both output duration control and extrapolation. VoiceStar is an autoregressive encoder-decoder neural codec language model, that leverages a novel Progress-Monitoring Rotary Position Embedding (PM-RoPE) and is trained with Continuation-Prompt Mixed (CPM) training. PM-RoPE enables the model to better align text and speech tokens, indicates the target duration for the generated speech, and also allows the model to generate speech waveforms much longer in duration than those seen during. CPM training also helps to mitigate the training/inference mismatch, and significantly improves the quality of the generated speech in terms of speaker similarity and intelligibility. VoiceStar outperforms or is on par with current state-of-the-art models on short-form benchmarks such as Librispeech and Seed-TTS, and significantly outperforms these models on long-form/extrapolation benchmarks (20-50s) in terms of intelligibility and naturalness. Code and models: https://github.com/jasonppy/VoiceStar. Audio samples: https://jasonppy.github.io/VoiceStar_web