Video to surface normals-map with NormalCrafter
retained text
I designed a couple of ComfyUI nodes to convert video to surface normals-map estimation using the model NormalCrafter This is very useful for 3D reconstruction, relighting, retexturing, video editing, and reference to video on H3 Minimax and Seedance I included a Normal temporal stabilizer, a complementary post-processing node that uses optical flow from the source video to reduce residual flicker while preserving real motion and changes in geometry. Processing is fast and run on a toaster (I included decode in chunks) Nodes and Workflow: [github ComfyUI-VizClick-NormalCrafter](https://github.com/zuanfilm/ComfyUI-VizClick-NormalCrafter) Models weights : [Huggingface NormalCrafter ](https://huggingface.co/Yanrui95/NormalCrafter/tree/main) arxiv Paper: [NormalCrafter: Learning Temporally Consistent Normals from Video Diffusion Priors](https://arxiv.org/pdf/2504.11427)
document fetched 2026-10-15 00:34 · canonical https://reddit.com/r/StableDiffusion/comments/1x65o2g/video_to_surface_normalsmap_with_normalcrafter