OpenImagingLab/FlashVSR

[CVPR 2026] Towards Real-Time Diffusion-Based Streaming Video Super-Resolution — An efficient one-step diffusion framework for streaming VSR with locality-constrained sparse attention and a tiny conditional decoder.

47
/ 100
Emerging

Combines one-step diffusion with locality-constrained sparse attention and a tiny conditional decoder to achieve ~17 FPS on 768×1408 video on a single A100 GPU. Employs a three-stage distillation pipeline enabling streaming inference while maintaining quality across ultra-high resolutions. Integrates with ComfyUI, HuggingFace model hub, and multiple cloud inference platforms, with a new VSR-120K dataset (120k videos, 180k images) supporting large-scale training.

1,430 stars.

No Package No Dependents
Maintenance 6 / 25
Adoption 10 / 25
Maturity 13 / 25
Community 18 / 25

How are scores calculated?

Stars

1,430

Forks

119

Language

Python

License

Apache-2.0

Last pushed

Dec 23, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/OpenImagingLab/FlashVSR"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.