README.md

metadata

title: FaceLift
emoji: 🌖
colorFrom: red
colorTo: pink
sdk: gradio
sdk_version: 5.49.1
app_file: app.py
pinned: false
license: cc-by-nc-sa-4.0

FaceLift: Single Image 3D Face Reconstruction

Transform a single portrait image into a complete 3D head model using multi-view diffusion and Gaussian Splatting.

Features

• Single Image Input: Upload any portrait photo
• Automatic Face Detection: Optional auto-cropping and alignment
• Multi-view Generation: Creates 6 consistent views using diffusion models
• 3D Reconstruction: Generates high-quality 3D Gaussian splats
• Interactive 3D Viewer: WebGL-based viewer for exploring the model (based on antimatter15/splat)
• Turntable Animation: Exports rotating 360° video
• Downloadable Model: Get the 3D model as a .ply file

Usage

1. Upload a portrait image
2. Adjust parameters (optional):
   • Auto Cropping: Enable for automatic face detection
   • Guidance Scale: Controls generation quality (default: 3.0)
   • Random Seed: For reproducible results
   • Generation Steps: Higher = better quality but slower
3. Click Submit and wait for processing
4. View outputs:
   • Download the Interactive 3D Viewer HTML file and open it in your browser for full-screen exploration
   • Watch the turntable animation
   • Download the 3D model (.ply file) for use in other software

Citation

@article{facelift2025,
  title={FaceLift: Single Image 3D Face Reconstruction},
  author={FaceLift Research Group},
  year={2025}
}

License

This software is free for non-commercial, research and evaluation use under the CC-BY-NC-SA-4.0 license.

For commercial use inquiries, contact: [email protected]