SeqTex ComfyUI Wrapper

VAST-AI SeqTex: Generate Mesh Textures in Video Sequence

(Arxiv Paper) (HuggingFace)

New Features!!!

• 10.11.2025: Added SeqTex Project Texture node for re-projecting edited multiview images back to UV space without running the WAN generator
• 20.08.2025: Fixed seam bleed in normal and position map bake
• 19.08.2025: Added support for Multi-view Image Input including SDXL workflow example

Installation

git clone https://github.com/Rizzlord/ComfyUI-SeqTex
[path_to_python_embeded]/python.exe -m pip install -r requirements.txt

SeqTex Project Texture

Need to skip the WAN texture generation and simply bake your edited multiview renders back onto the mesh?
Hook the SeqTex Step 1 mesh plus its mvp_matrix / w2c tensor paths and your multiview RGB batch into SeqTex Project Texture.
The node replays the exact camera rig with nvdiffrast, handles preset-based view trimming (2/4/6/12), optional per-view masks, and outputs a UV texture tensor you can save or feed downstream.

Authors

Rizzlord easymode