---
license: apache-2.0
language:
- en
- zh
- es
- fr
- de
pipeline_tag: audio-to-audio
---
# UniPASE

UniPASE is a state-of-the-art generative model for universal speech enhancement (USE), with its paper accepted by *IEEE TASLP*.

## Model Sources
- **Repository**: https://github.com/Xiaobin-Rong/unipase
- **Paper**: https://arxiv.org/abs/2604.14606
- **Demo**: https://xiaobin-rong.github.io/unipase_demo/