InfiniteYou is a robust framework developed by ByteDance for identity-preserving image generation. Leveraging Diffusion Transformers (DiTs), specifically FLUX, it addresses challenges like insufficient identity similarity and poor text-image alignment. Central to InfiniteYou is InfuseNet, a component that injects identity features into the DiT base model via residual connections, enhancing identity similarity while maintaining generation capabilities. A multi-stage training strategy, including pretraining and supervised fine-tuning with synthetic single-person-multiple-sample (SPMS) data, further improves text-image alignment and image quality.
Key Features
Utilizes InfuseNet to integrate identity features into the DiT model.
Employs a multi-stage training strategy with pretraining and supervised fine-tuning.
Supports two model variants: aes_stage2 for better text-image alignment and aesthetics, and sim_stage1 for higher identity similarity.
Plug-and-play design ensures compatibility with various existing methods.
Official ComfyUI node available for seamless integration
Android DevHub
.png)
.png)
.png)
