See-through: Single-image Layer Decomposition for Anime Characters

TL;DR

This paper presents a novel framework that automatically decomposes a single anime illustration into manipulatable layers, enabling efficient creation of 2.5D models for animation without manual segmentation.

Contribution

It introduces a scalable engine for high-quality supervision from Live2D models and combines diffusion-based and pseudo-depth methods for detailed layer decomposition.

Findings

Achieves high-fidelity, manipulatable anime models

Enables real-time animation workflows

Overcomes manual segmentation limitations

Abstract

We introduce a framework that automates the transformation of static anime illustrations into manipulatable 2.5D models. Current professional workflows require tedious manual segmentation and the artistic ``hallucination'' of occluded regions to enable motion. Our approach overcomes this by decomposing a single image into fully inpainted, semantically distinct layers with inferred drawing orders. To address the scarcity of training data, we introduce a scalable engine that bootstraps high-quality supervision from commercial Live2D models, capturing pixel-perfect semantics and hidden geometry. Our methodology couples a diffusion-based Body Part Consistency Module, which enforces global geometric coherence, with a pixel-level pseudo-depth inference mechanism. This combination resolves the intricate stratification of anime characters, e.g., interleaving hair strands, allowing for dynamic layer reconstruction. We demonstrate that our approach yields high-fidelity, manipulatable models suitable for professional, real-time animation applications.