see-through-anime-layer-decomposition

Skill for See-through (SIGGRAPH 2026 conditionally accepted) — a diffusion-based framework that decomposes a single anime illustration into up to 23 inpainted, depth-ordered semantic layers and exports a 2.5D-ready layered PSD. Walks through conda install, HuggingFace models (LayerDiff 3D, Marigold, SAM body parsing), `inference_psd.py`, and heuristic depth/LR splits.