---
license: cc-by-sa-4.0
pipeline_tag: text-to-audio
---

# VoxAudio

VoxAudio is a streaming chunk-autoregressive flow matching model for vocalized audio generation: given a text caption that may quote explicit speech, it generates 24 kHz audio containing articulate speech together with the surrounding soundscape.

- **Paper**: [VoxAudio: Vocalized Audio Synthesis via Multi-Reward Autoregressive Flow Matching](https://huggingface.co/papers/2608.12951)
- **Project Page**: [https://voxaudio.github.io](https://voxaudio.github.io)
- **Code**: [https://github.com/VoxAudio/VoxAudio_Main](https://github.com/VoxAudio/VoxAudio_Main)

For setup and inference instructions, please refer to the GitHub repository.