From ad3e92e6362d30902db5c223aa119ed2e17df1d6 Mon Sep 17 00:00:00 2001 From: smthemex <138738845+smthemex@users.noreply.github.com> Date: Wed, 5 Nov 2025 22:07:48 +0800 Subject: [PATCH] Update README.md --- README.md | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/README.md b/README.md index 87fbcc6..20dc1bb 100644 --- a/README.md +++ b/README.md @@ -2,6 +2,7 @@ [FlashVSR](https://github.com/OpenImagingLab/FlashVSR): Towards Real-Time Diffusion-Based Streaming Video Super-Resolution,this node ,you can use it in comfyUI # Upadte +* update to version v1.1 /更新适配1.1版本的新模型和代码,降低闪烁,提高保真度和稳定性 * add full mode [lightx2v vae encoder](https://huggingface.co/lightx2v/Autoencoders/tree/main) support(only lightvaew2_1.pth,taew2_1.pth,lighttaew2_1.pth) and [Wan2.1-VAE-upscale2x](https://huggingface.co/spacepxl/Wan2.1-VAE-upscale2x) support * 新增lightx2v 加速vae decoder支持和Wan2.1-VAE-upscale2x 放大decoder支持,只是在full 模式下有效,light的加速模型目前只支持(lightvaew2_1.pth #32.2M,taew2_1.pth,lighttaew2_1.pth) 三个文件 # Tips @@ -12,7 +13,7 @@ * local_range=7这个是会最清晰,local_range=11会比较稳定,color fix 推荐用小波(没重影); * 编译Block-Sparse-Attention window的轮子 可以使用 [ smthemex 强制编译版](https://github.com/smthemex/Block-Sparse-Attention) 或者 [lihaoyun6 要联网](https://github.com/lihaoyun6/Block-Sparse-Attention) 两个fork来,不推荐用官方的 * Block-Sparse-Attention 正确安装且能调用才是方法的完全体,当前的函数实现会更容易OOM,但是Block-Sparse-Attention轮子实在不好找,目前只有[CU128 toch2.7](https://github.com/lihaoyun6/ComfyUI-WanVideoWrapper)的,我提供的([cu128,torch2.8,py311单体](https://pan.quark.cn/s/c9ba067c89bc))或者自己编译 -* 方法是基于现有prompt.pt训练的,所以外置cond没有必要已经去掉,新增tile 和 color fix 选项,tile关闭质量更高,需要VRam更高,corlor fix对于非模糊图片可以试试。修复图片索引数不足的错误。 +* 方法是基于现有prompt.pt训练的,新增tile 和 color fix 选项,tile关闭质量更高,需要VRam更高,corlor fix对于非模糊图片可以试试。修复图片索引数不足的错误。 * Choice vae infer full mode ,encoder infer tiny mode 选择vae跑full模式 效果最好,tiny则是速度,数据集基于4倍训练,所以1 scale是不推荐的; * 如果觉得项目有用,请给官方项目[FlashVSR](https://github.com/OpenImagingLab/FlashVSR) 打星; if you Like it , star the official project [link](https://github.com/OpenImagingLab/FlashVSR) @@ -45,15 +46,16 @@ python setup.py install 3.checkpoints ---- -* 3.1 [FlashVSR](https://huggingface.co/JunhaoZhuang/FlashVSR/tree/main) all checkpoints 所有模型,vae 用常规的wan2.1 +* 3.1.2 [FlashVSRv1.0](https://huggingface.co/JunhaoZhuang/FlashVSR/tree/main) all checkpoints 所有模型,vae 用常规的wan2.1 +* 3.1.2 [FlashVSRv1.1](https://huggingface.co/JunhaoZhuang/FlashVSR-v1.1/tree/main) all checkpoints 所有模型,vae 用常规的wan2.1 * 3.2 emb [posi_prompt.pth](https://github.com/OpenImagingLab/FlashVSR/tree/main/examples/WanVSR/prompt_tensor) 4M而已 * 3.3 [lightvaew2_1.pth](https://huggingface.co/lightx2v/Autoencoders/tree/main) and [diffusion_pytorch_model.safetensors](https://huggingface.co/spacepxl/Wan2.1-VAE-upscale2x/tree/main/diffusers/Wan2.1_VAE_upscale2x_imageonly_real_v1) ``` ├── ComfyUI/models/FlashVSR -| ├── LQ_proj_in.ckpt +| ├── LQ_proj_in.ckpt # v1.1 or v1.0 | ├── TCDecoder.ckpt -| ├── diffusion_pytorch_model_streaming_dmd.safetensors +| ├── diffusion_pytorch_model_streaming_dmd.safetensors #v1.1 or v1.0 | ├── posi_prompt.pth ├── ComfyUI/models/vae | ├──Wan2.1_VAE.pth