lootingTree / README.md
VoidGx's picture
Update README.md
fe30c6c verified
|
Raw History Blame Contribute Delete
2.52 kB

AI Manga/Manhwa Parallax Animator

Description

AI-powered tool to create dynamic parallax animations from manga or manhwa artwork. The tool analyzes static illustrations and generates cinematic motion effects β€” parallax layers, camera zooms, shakes, and style-preserving movements β€” and can export sequences compatible with After Effects for further refinement.

Key Features

  • AI-Powered Motion Analysis: layer detection, matting, and depth estimation stubs provided
  • Dynamic Zoom & Shake Effects: camera-path and procedural shake modules (stubs)
  • Style Preservation: designed to keep original line art and halftone textures
  • After Effects Integration: JSON export template and exporter class
  • CLI for quick iteration and example demo

Quickstart

  1. Clone the repository
  2. Create a virtualenv and install dependencies: pip install -r requirements.txt
  3. Run the demo (replace input.png with your artwork): python -m animator.cli --input examples/sample_art.png --output out.mp4 --export-ae out_ae.json

Project layout

  • README.md β€” this file
  • LICENSE β€” MIT license
  • requirements.txt β€” minimal Python deps
  • .gitignore
  • src/animator/ β€” core Python package
    • core.py β€” processing pipeline and primary classes
    • ae_exporter.py β€” After Effects export helper
  • scripts/cli.py β€” command-line entrypoint
  • examples/demo.py β€” demo pipeline runner
  • aftereffects/export_template.json β€” AE-friendly JSON structure template

Architecture (high-level)

  1. Input image -> Preprocessing (resize, normalize)
  2. Layer & matte extraction (segmentation / matting models)
  3. Depth / parallax map estimation
  4. Motion plan generation (paths, camera zooms, shakes)
  5. Frame synthesis / compositing (render frames with transforms + optional style-preservation passes)
  6. Export video (MP4 / WebM) and After Effects JSON for further refinement

Extending the scaffold

  • Replace stubbed implementations in src/animator/core.py with your chosen ML models (e.g., U^2-Net for matting, MiDaS or custom depth estimator).
  • Implement a GPU-accelerated renderer that composes per-layer transforms and output frames.
  • Expand ae_exporter.py to match your After Effects scripting workflow or JSON schema expected by third-party AE importers.

License

MIT. See LICENSE file.

Contributing

Pull requests and issues are welcome. For major changes, open an issue first to discuss the plan.