shitagaki-lab

shitagaki-lab / see-through

Public

"Single-image Layer Decomposition for Anime Characters" (SIGGRAPH 2026, Conditionally Accepted)

682
62
100% credibility
Found Apr 01, 2026 at 670 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Jupyter Notebook
AI Summary

An open-source tool that decomposes a single anime illustration into up to 23 inpainted semantic layers with depth ordering for easy PSD export and 2.5D manipulation.

How It Works

1
🔍 Discover See-Through

You find this fun tool online that turns flat anime drawings into editable layers.

2
🖼️ Pick your anime image

Choose a favorite anime character picture from your photos.

3
🚀 Open the app

Launch the simple app and drop in your image.

4
✨ Decompose into layers

Click go and it magically splits your character into 23 smart parts like hair, eyes, and clothes with perfect fills.

5
👀 Preview the magic

See each layer separately, ready to move and edit.

🎉 Export and create

Save the layered file and start animating your character!

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 670 to 682 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is see-through?

See-through takes a single anime character image and decomposes it into up to 23 semantic layers—like hair, face, eyes, clothing, and accessories—with automatic inpainting of occluded areas and inferred depth-based drawing order. It outputs ready-to-use layered PSD files via a simple CLI: `python inference/scripts/inference_psd.py --srcp image.png --save_to_psd`. Built in Jupyter Notebooks with PyTorch diffusion models for single image depth estimation and layer generation, plus SAM for segmentation, it turns flat illustrations into editable 2.5D stacks.

Why is it gaining traction?

Unlike generic single image layer separation tools, see-through nails anime specifics like see-through bangs and intricate outfits, automating what used to take hours of manual Photoshop work. The HF Space demo and ComfyUI integration let devs test see-through views instantly, while heuristic scripts refine layers by depth or left-right splits. At 537 stars, it's hooking the niche for single image to 3D workflows without retraining.

Who should use this?

Live2D riggers prepping models from fanart, anime game devs needing quick single layer image exports for Unity see-through shaders, or researchers prototyping single image portrait relighting and super-resolution pipelines. Skip if you're not in anime—it's tuned for that style.

Verdict

Grab it for anime layer decomp; the SIGGRAPH 2026 acceptance signals quality despite 1.0% credibility score and pending training docs. 537 stars show momentum, but expect research-grade maturity—test on your assets first.

(187 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.