A comprehensive zero-loss pipeline for transforming single-view food photos into optimized, production-ready 3D models using ComfyUI Gaussian Splatting, SuperSplat filtering, KIRI Engine shading, and Blender AI Agent automation.
I have imported a food model generated via Gaussian Splatting, but it contains messy, unfinished fragments of the original white plate. Please execute the following automation steps via MCP: 1. Analyze the imported model mesh. Identify and select the jagged, flat, unfinished "plate shards" located at the very bottom base of the food (surrounding the bowl and the bread). 2. Delete these messy plate fragments/vertices completely, leaving only the main food elements (the stew bowl, the bread, and the lettuce). 3. Create a brand-new, clean, perfectly circular 3D cylinder/disc mesh to act as a ceramic plate directly underneath the remaining food elements. 4. Position and scale the new plate so the food elements sit naturally inside its upper lip. 5. Create a new material named "Ceramic_White". Set its base color to pure white, increase the Specular to 0.6, and lower the Roughness to 0.15 to give it a realistic glossy ceramic look. Apply this material to the new plate mesh. 6. Select the bottom vertices of the food elements, and slightly flatten or sculpt them down so they cleanly intersect with the surface of the new plate without clipping through the bottom.