Downloads · 30 days
0
giahuy102/FlashVSR
FlashVSR is a machine learning model from giahuy102. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
This repository hosts the FlashVSR model weights used in the ComfyUI-FlashVSR project. It provides fast and efficient video super-resolution powered by FlashAttention-based architectures.
Downloads · 30 days
0
Access
Public
Updated Feb 27, 2026
Repo size
14.2 GB
Likes
0
Public
Click a slice to open those files.
.safetensors18.6 GB · 94%
From the Hugging Face model README
This repository hosts the FlashVSR model weights used in the ComfyUI-FlashVSR project. It provides fast and efficient video super-resolution powered by FlashAttention-based architectures.
FlashVSR (Flash Video Super-Resolution) is a high-performance video SR model that leverages efficient attention computation to upscale low-resolution videos while maintaining temporal consistency.
This version is provided for ComfyUI integration, optimized for direct use with the ComfyUI-FlashVSR extension.
| File | Description | Size |
|---|---|---|
Wan2_1-T2V-1_3B_FlashVSR_fp32.safetensors | Main FlashVSR model weights | ~5.7 GB |
Wan2_1_FlashVSR_TCDecoder_fp32.safetensors | Temporal consistency decoder | ~181 MB |
Wan2_1_FlashVSR_LQ_proj_model_bf16.safetensors | Low-quality projection model | ~576 MB |
Wan2.1_VAE.safetensors | VAE weights | ~254 MB |
Prompt.safetensors | Conditioning embeddings | ~4 MB |
Install ComfyUI
Clone the FlashVSR node extension:
git clone https://github.com/1038lab/ComfyUI-FlashVSR.git
Download the model files from this Hugging Face repo:
Place them inside:
ComfyUI/models/FlashVSR/
Launch ComfyUI and load the FlashVSR workflow.
This model is based on the research and implementation from 👉 JunhaoZhuang/FlashVSR All credit for the model architecture and original training goes to the authors.
This repository only repackages the model for ComfyUI integration.
Please refer to the original FlashVSR license for usage rights. Any redistribution or fine-tuning should comply with the same terms.