See Through Anime Layer Decomposition

This breaks down a single anime illustration into up to 23 separate, fully inpainted layers with depth ordering, then exports everything as a PSD ready for 2.5D animation workflows. Built on SDXL and a fine-tuned Marigold depth model, it handles semantic splits like hair, face, eyes, clothing, and accessories while completing occluded regions behind each part. You'd use this when you need manipulatable layers from flat art without manually masking everything. The heuristic post-processing scripts let you further stratify layers by depth or left-right position. Installation is heavy with optional tiers for different annotators, and you need to run everything from repo root. The result is production-ready layered files, not just segmentation masks.

Install to Claude Code

npx -y skills add aradotso/trending-skills --skill see-through-anime-layer-decomposition --agent claude-code