See-through turns anime images into editable 2.5D layers
See-through is an open-source research project that decomposes a single anime illustration into inpainted, semantically distinct layers with inferred drawing order. The released pipeline can produce PSD output with up to 23 layers and includes low-VRAM options for GPUs with 8 GB or 12 GB of memory. The project is presented as a SIGGRAPH 2026 conference paper and also includes training scripts, model configurations, data pipelines, and inference tools.
See-through provides downloadable code, models, and workflows for automating anime segmentation and occluded-region inpainting, which the maintainers identify as useful starting points for Live2D artwork.
"Single-image Layer Decomposition for Anime Characters" (SIGGRAPH 2026 Conference Paper)
See-through: Single-image Layer Decomposition for Anime Characters
Jian Lin1, Chengze Li1*, Haoyun Qin2,3,4, Kwun Wang Chan1, Yanghua Jin3, Hanyuan Liu1, Stephen Chun Wang Choy1, Xueting Liu1
1Saint Francis University 2University of Pennsylvania 3Spellbrush 4Shitagaki Lab
*Corresponding author
#See-through#LayerDiff 3D#Marigold Depth#SAM 2#Shitagaki Lab