EXPRESSO: A Benchmark and Analysis of Discrete Expressive Speech Resynthesis

TL;DR

EXPRESSO tackles the scarcity of expressive, textless speech data for neural resynthesis by introducing a 47-hour expressive dataset with read and improvised dialogues across 26 styles. It analyzes discrete resynthesis using self-supervised HuBERT-based units and Encodec codebooks, paired with vocoders conditioned on speaker and style, and evaluates content, style, and pitch preservation. Key findings show Encodec-based units deliver strong resynthesis quality and pitch/style preservation, while HuBERT-based units offer better content fidelity and controllability, with some cross-domain style transfer challenges. The dataset and evaluation framework are open source to spur further development in expressive, textless speech synthesis.

Abstract

Recent work has shown that it is possible to resynthesize high-quality speech based, not on text, but on low bitrate discrete units that have been learned in a self-supervised fashion and can therefore capture expressive aspects of speech that are hard to transcribe (prosody, voice styles, non-verbal vocalization). The adoption of these methods is still limited by the fact that most speech synthesis datasets are read, severely limiting spontaneity and expressivity. Here, we introduce Expresso, a high-quality expressive speech dataset for textless speech synthesis that includes both read speech and improvised dialogues rendered in 26 spontaneous expressive styles. We illustrate the challenges and potentials of this dataset with an expressive resynthesis benchmark where the task is to encode the input in low-bitrate units and resynthesize it in a target voice while preserving content and style. We evaluate resynthesis quality with automatic metrics for different self-supervised discrete encoders, and explore tradeoffs between quality, bitrate and invariance to speaker and style. All the dataset, evaluation metrics and baseline models are open source