ASAICLLGSDMar 6, 2025

Scaling Rich Style-Prompted Text-to-Speech Datasets

arXiv:2503.04713v222 citationsh-index: 16Has CodeEMNLP
Originality Incremental advance
AI Analysis

This addresses the problem of limited rich style annotations for TTS researchers, though it is incremental as it builds on existing methods for scaling annotations.

The authors tackled the lack of large-scale datasets with rich style captions for text-to-speech by introducing ParaSpeechCaps, a dataset combining human-labelled and automatically annotated data, which improved style consistency by 7.9% and speech quality by 15.5% when used to finetune a TTS model.

We introduce Paralinguistic Speech Captions (ParaSpeechCaps), a large-scale dataset that annotates speech utterances with rich style captions. While rich abstract tags (e.g. guttural, nasal, pained) have been explored in small-scale human-annotated datasets, existing large-scale datasets only cover basic tags (e.g. low-pitched, slow, loud). We combine off-the-shelf text and speech embedders, classifiers and an audio language model to automatically scale rich tag annotations for the first time. ParaSpeechCaps covers a total of 59 style tags, including both speaker-level intrinsic tags and utterance-level situational tags. It consists of 342 hours of human-labelled data (PSC-Base) and 2427 hours of automatically annotated data (PSC-Scaled). We finetune Parler-TTS, an open-source style-prompted TTS model, on ParaSpeechCaps, and achieve improved style consistency (+7.9% Consistency MOS) and speech quality (+15.5% Naturalness MOS) over the best performing baseline that combines existing rich style tag datasets. We ablate several of our dataset design choices to lay the foundation for future work in this space. Our dataset, models and code are released at https://github.com/ajd12342/paraspeechcaps .

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes