VACE Subject Replace! Replace Anything in Videos with Wan2.1+VACE

Use this workflow to replace any subject in a video with Wan2.1+VACE. It takes a video and produces a video with the subject replaced.

The graph includes VHS_LoadVideo, Sam2Segmentation, WanVideoVACEEncode, WanVideoSampler, WanVideoDecode, and VHS_VideoCombine nodes.

The graph provides VHS_LoadVideo and LoadImage for inputs, with VHS_VideoCombine, PreviewImage, and SaveImage for output handling.

The listed base model is Wan Video.

The workflow lists comfyroll, comfyui_controlnet_aux, comfyui_essentials, and comfyui_layerstyle as node packs. It also lists comfyui-kjnodes, comfyui-videohelpersuite, ComfyUI-WanVideoWrapper, and segment-anything-2.

Listed model files include flux/flux_vae.sft, flux/flux1-dev.sft, instantX-Flux-Union.safetensors, and juggernautXL_juggXIByRundiffusion.safetensors. The list also includes sai_xl_depth_256lora.safetensors, sam2.1_hiera_large.safetensors, wan/Wan2_1_VACE_1_3B_preview_bf16.safetet

The remaining listed model files are wan/Wan2_1-T2V-1_3B_bf16.safetensors and wan/wrapper_Wan2_1_VAE_fp32.safetensors.

Consider incorporating depth, openpose, or inverted lineart instead of using only a gray background.

Suggestion · not verified

The graph includes DepthAnythingV2Preprocessor, DWPreprocessor, and LineArtPreprocessor.

Before use, verify which package provides comfyui_controlnet_aux and whether wan/wrapper_Wan2_1_VAE_fp32.safetensors is available under that exact name.

Suggestion · not verified