Leffa: Learning Flow Fields in Attention for Controllable Person Image Generation

📚 Paper - 🤖 Code - 🔥 Demo - 🤗 Model

Star ⭐ us if you like it!

News

  • 09/Jan/2025. Inference defaults to float16, generating an image in 6 seconds (on A100).

More news can be found in the GitHub repository.

Leffa is a unified framework for controllable person image generation that enables precise manipulation of both appearance (i.e., virtual try-on) and pose (i.e., pose transfer).

Person Image
Examples

Garment Image
Examples

Generated Image
Model Type
Garment Type
Accelerate Reference UNet (may slightly reduce performance)
Repaint Mode

Note: The models used in the demo are trained solely on academic datasets. Virtual try-on uses VITON-HD/DressCode, and pose transfer uses DeepFashion.