Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2509.21522 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916970749755392 |
|---|---|
| author | Zhou, Naisong Phaye, Saisamarth Rajesh Cernak, Milos Stojkovic, Tijana Pearce, Andy Cavallaro, Andrea Harper, Andy |
| author_facet | Zhou, Naisong Phaye, Saisamarth Rajesh Cernak, Milos Stojkovic, Tijana Pearce, Andy Cavallaro, Andrea Harper, Andy |
| contents | Diffusion-based generative models have achieved state-of-the-art performance for perceptual quality in speech enhancement (SE). However, their iterative nature requires numerous Neural Function Evaluations (NFEs), posing a challenge for real-time applications. On the contrary, flow matching offers a more efficient alternative by learning a direct vector field, enabling high-quality synthesis in just a few steps using deterministic ordinary differential equation~(ODE) solvers. We thus introduce Shortcut Flow Matching for Speech Enhancement (SFMSE), a novel approach that trains a single, step-invariant model. By conditioning the velocity field on the target time step during a one-stage training process, SFMSE can perform single, few, or multi-step denoising without any architectural changes or fine-tuning. Our results demonstrate that a single-step SFMSE inference achieves a real-time factor (RTF) of 0.013 on a consumer GPU while delivering perceptual quality comparable to a strong diffusion baseline requiring 60 NFEs. This work also provides an empirical analysis of the role of stochasticity in training and inference, bridging the gap between high-quality generative SE and low-latency constraints. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_21522 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Shortcut Flow Matching for Speech Enhancement: Step-Invariant flows via single stage training Zhou, Naisong Phaye, Saisamarth Rajesh Cernak, Milos Stojkovic, Tijana Pearce, Andy Cavallaro, Andrea Harper, Andy Sound Artificial Intelligence Diffusion-based generative models have achieved state-of-the-art performance for perceptual quality in speech enhancement (SE). However, their iterative nature requires numerous Neural Function Evaluations (NFEs), posing a challenge for real-time applications. On the contrary, flow matching offers a more efficient alternative by learning a direct vector field, enabling high-quality synthesis in just a few steps using deterministic ordinary differential equation~(ODE) solvers. We thus introduce Shortcut Flow Matching for Speech Enhancement (SFMSE), a novel approach that trains a single, step-invariant model. By conditioning the velocity field on the target time step during a one-stage training process, SFMSE can perform single, few, or multi-step denoising without any architectural changes or fine-tuning. Our results demonstrate that a single-step SFMSE inference achieves a real-time factor (RTF) of 0.013 on a consumer GPU while delivering perceptual quality comparable to a strong diffusion baseline requiring 60 NFEs. This work also provides an empirical analysis of the role of stochasticity in training and inference, bridging the gap between high-quality generative SE and low-latency constraints. |
| title | Shortcut Flow Matching for Speech Enhancement: Step-Invariant flows via single stage training |
| topic | Sound Artificial Intelligence |
| url | https://arxiv.org/abs/2509.21522 |