| title | One Model, Many Latencies: Universal Speech Enhancement for Diverse Real-Time Applications |
|---|---|
| authors | Szu-Wei Fu, Rong Chao, Xuesong Yang, Sung-Feng Huang, Ante Jukić, Yu Tsao, Yu-Chiang Frank Wang |
| date | 2026-06-24 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2606.25621 |
| affiliation | NVIDIA |
| upvotes | 5 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf |
| abstract | Different real-time speech applications impose distinct latency budgets, often requiring separately trained enhancement models for each scenario. In this paper, we propose a one-for-all, real-time universal speech enhancement model that provides explicit control over both algorithmic and computational latency. Algorithmic latency is flexibly adjusted via configurable look-ahead frames. To avoid learning inefficiency caused by varying padding configurations, we introduce parallel convolutional layers corresponding to different look-ahead settings. Computational latency is controlled through an early-exit mechanism, enabling inference at different network depths. To narrow the performance gap between specialized and flexible models, we propose a two-stage training strategy with a shared-to-multiple decoder transition. Overall, the proposed framework enables a single model to be deployed across diverse latency budgets without retraining separate models. |
cite
@misc{fu2026one,
title={One Model, Many Latencies: Universal Speech Enhancement for Diverse Real-Time Applications},
author={Szu-Wei Fu and Rong Chao and Xuesong Yang and Sung-Feng Huang and Ante Jukić and Yu Tsao and Yu-Chiang Frank Wang},
year={2026},
eprint={2606.25621},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2606.25621}
}