CapSpeech: Enabling Downstream Applications in Style-Captioned Text-to-Speech
Abstract
CapSpeech introduces a large benchmark dataset for various captioned text-to-speech tasks, facilitating advancements in style, accent, emotion, and chat-agent synthesis.
Recent advancements in generative artificial intelligence have significantly transformed the field of style-captioned text-to-speech synthesis (CapTTS). However, adapting CapTTS to real-world applications remains challenging due to the lack of standardized, comprehensive datasets and limited research on downstream tasks built upon CapTTS. To address these gaps, we introduce CapSpeech, a new benchmark designed for a series of CapTTS-related tasks, including style-captioned text-to-speech synthesis with sound events (CapTTS-SE), accent-captioned TTS (AccCapTTS), emotion-captioned TTS (EmoCapTTS), and text-to-speech synthesis for chat agent (AgentTTS). CapSpeech comprises over 10 million machine-annotated audio-caption pairs and nearly 0.36 million human-annotated audio-caption pairs. In addition, we introduce two new datasets collected and recorded by a professional voice actor and experienced audio engineers, specifically for the AgentTTS and CapTTS-SE tasks. Alongside the datasets, we conduct comprehensive experiments using both autoregressive and non-autoregressive models on CapSpeech. Our results demonstrate high-fidelity and highly intelligible speech synthesis across a diverse range of speaking styles. To the best of our knowledge, CapSpeech is the largest available dataset offering comprehensive annotations for CapTTS-related tasks. The experiments and findings further provide valuable insights into the challenges of developing CapTTS systems.
Community
We are excited to share our recent work titled "CapSpeech: Enabling Downstream Applications in Style-Captioned Text-to-Speech"
📄 Paper: https://arxiv.org/abs/2506.02863
🌐 Project Page: https://wanghelin1997.github.io/CapSpeech-demo/
🚀 Spaces Demo: https://huggingface.co/spaces/OpenSound/CapSpeech-TTS
CapSpeech is a new benchmark designed for style-captioned TTS (CapTTS) tasks, including style-captioned text-to-speech synthesis with sound effects (CapTTS-SE), accent-captioned TTS (AccCapTTS), emotion-captioned TTS (EmoCapTTS) and text-to-speech synthesis for chat agent (AgentTTS).
CapSpeech comprises over 10 million machine-annotated audio-caption pairs and nearly 0.36 million human-annotated audio-caption pairs. 3 new speech datasets are specifically designed for the CapTTS-SE and AgentTTS tasks to enhance the benchmark’s coverage of real-world scenarios.
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- Benchmarking Expressive Japanese Character Text-to-Speech with VITS and Style-BERT-VITS2 (2025)
- Muyan-TTS: A Trainable Text-to-Speech Model Optimized for Podcast Scenarios with a $50K Budget (2025)
- Kimi-Audio Technical Report (2025)
- AlignDiT: Multimodal Aligned Diffusion Transformer for Synchronized Speech Generation (2025)
- Shallow Flow Matching for Coarse-to-Fine Text-to-Speech Synthesis (2025)
- GSA-TTS : Toward Zero-Shot Speech Synthesis based on Gradual Style Adaptor (2025)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment:
@librarian-bot
recommend
Models citing this paper 1
Datasets citing this paper 14
Browse 14 datasets citing this paperSpaces citing this paper 2
Collections including this paper 0
No Collection including this paper