Object detection
Finds buttons, frames, icons, bars, cards, currencies, and UI fragments from a single flattened image.
Feature set
Extraction tools for people who still inspect the output.
CakeSplit combines automated detection with a visual review surface, so production teams can trust what they export.
Clean masks
Extracts transparent assets with edge-aware matting for glows, bevels, soft shadows, and irregular silhouettes.
Layer workspace
Review, rename, lock, hide, group, and reorder assets before they leave the browser.
Style variations
Generate alternate art treatments for a selected asset while keeping the rest of the screen stable.
Manual correction
Refine detections with selection and proofing tools when a production screenshot needs extra care.
Export bundles
Send assets into Figma, Photoshop, Unity, Godot, or plain PNG/JSON workflows.
AI-assisted detection, matting, and style generation can miss details or create artifacts. CakeSplit is designed around review before export.
Exports that keep context.
Every export includes names, bounds, and descriptors where the target tool supports it. The goal is a usable handoff, not just a folder of images.
Figma
Component tree + placement metadata
Photoshop
JSX automation script
Unity
ZIP with engine-ready metadata
Godot
PNG assets and JSON descriptor
Ready for a real screenshot
Test the feature set on your hardest screenshot.
Start free with 100 credits and see exactly what CakeSplit detects.
Start freeinput.png12 layers detected
HUD
Button
Badge
Coin
Frame
Glow