和上篇一样,由于wan2.2的发布,更新了下工作流,同样分为t2v和i2v两个。生成效果相对上代有所提升。
工作流下载:wan2-video-v2 。
压缩包中同样包含2个工作流文件:
- Wan2.2 T2V(文生视频):wan2-t2v-upscale-v2.json
- Wan2.2 I2V(图生视频):wan2-i2v-upscale-v2.json
同样使用GGUF版量化模型运行,以节省资源。
以实际测试使用来说:具有8GB vRAM的RTX 4060显卡可使用 Q4_K_M 的主模型跑起来,具有 16GB vRAM的4070 Ti Super和4080 laptop显卡可采用 Q5_K_M 主模型运行,更高量化没有测试。
需求
模型:
Main组中:
- Wan 2.2 (GGUF), Put in: ComfyUI\models\unet
- Wan2.2-T2V
https://huggingface.co/QuantStack/Wan2.2-T2V-A14B-GGUF/resolve/main/HighNoise/Wan2.2-T2V-A14B-HighNoise-Q4_K_M.gguf
https://huggingface.co/QuantStack/Wan2.2-T2V-A14B-GGUF/resolve/main/LowNoise/Wan2.2-T2V-A14B-LowNoise-Q4_K_M.gguf - Wan2.2-I2V
https://huggingface.co/QuantStack/Wan2.2-I2V-A14B-GGUF/resolve/main/HighNoise/Wan2.2-I2V-A14B-HighNoise-Q4_K_M.gguf
https://huggingface.co/QuantStack/Wan2.2-I2V-A14B-GGUF/resolve/main/LowNoise/Wan2.2-I2V-A14B-LowNoise-Q4_K_M.gguf
- Wan2.2-T2V
- umt5-xxl-encoder-Q5_K_M.gguf (CLIP) Put in: ComfyUI\models\text_encoders
https://huggingface.co/city96/umt5-xxl-encoder-gguf/resolve/main/umt5-xxl-encoder-Q5_K_M.gguf - wan_2.1_vae.safetensors (VAE) Put in: ComfyUI\models\vae
https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors - 可选 Lora, 用于加速; Put in: ComfyUI\models\loras\LightX2V
页面地址:https://civitai.com/models/1838893?modelVersionId=2090458
由于Main处理的效果已经挺好了,如果显卡足够强,直接改输出分辨率即可,将Mian的输入接到后面的Interpolate进行插帧即可。
Upscaled组中:
- RealESRGAN_x2plus.pth or nmkdSiaxCX_200k.safetensors (Upscale Model) Put in: ComfyUI\models\upscale_models
https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.1/RealESRGAN_x2plus.pth
https://civitai.com/api/download/models/164677?type=Model&format=SafeTensor
经测nmkdSiaxCX_200k的效果更好一点。
V2V组中:
- 使用原有的wan2.1_t2v_1.3B with wan_2.1_vae 或新的 Wan2.2-TI2V-5B with wan2.2_vae 都可以
https://huggingface.co/QuantStack/Wan2.2-TI2V-5B-GGUF/resolve/main/Wan2.2-TI2V-5B-Q8_0.gguf
Put in: ComfyUI\models\unet
https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan2.2_vae.safetensors
Put in: ComfyUI\models\vae
或者:
https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors
Put in: ComfyUI\models\diffusion_models
wan_2.1_vae和Main组相同。 - umt5_xxl_fp8_e4m3fn_scaled.safetensors (CLIP, or same in “Main”) Put in: ComfyUI\models\text_encoders
ComfyUI Nodes:
- rgthree-comfy
- ComfyUI-KJNodes
- ComfyUI-VideoHelperSuite
- ComfyUI-Frame-Interpolation
- Comfyui-Memory_Cleanup
页码: 1 2