GenielabsOpenSource/spine-animation-ai

What it solves

Spine Animation AI automates the tedious and manual process of creating 2D skeletal animations in Spine. It removes the need for manual rigging, bone placement, and frame-by-frame animation for game characters, allowing creators to go from raw image assets to interactive, rigged animations quickly.

How it works

The project functions as a Claude agent skill that combines computer vision and AI-driven generation:

• Asset Positioning: Uses SIFT (Scale-Invariant Feature Transform) and RANSAC to automatically align separated body parts to a reference image. It also uses occlusion analysis to determine the correct draw order (Z-order).
• Rigging & Animation: Generates the necessary Spine JSON skeleton and bone hierarchies. It applies animation presets (like idle, walk, or jump) based on the 12 principles of animation using bezier easing.
• Reskinning: The "Reskin Studio" uses Gemini to repaint characters based on text descriptions, slicing the results back into textures for the existing rig.
• Character Splitting: For single images, it uses Gemini and OpenCV to deconstruct a character into individual transparent PNG parts.
• Previewing: Produces self-contained HTML5 previews using the official Spine Web Player.

Who it’s for

Game developers and 2D animators who use Spine 2D and want to automate the workflow of rigging and animating characters from static art assets.

Highlights

• Claude Integration: Can be added to Claude Projects via a single SKILL.md file, allowing the AI to run the pipeline directly.
• Automated Rigging: Auto-positions body parts and builds skeleton JSONs without manual placement.
• AI Reskinning: Change a character's appearance with a single sentence using Gemini.
• Animation Presets: Includes pre-built logic for common movements like walking, running, and attacking.
• Computer Vision Pipeline: Employs SIFT, RANSAC, and FLANN matchers for robust asset alignment.