✦
Summary
InfiniteYou is an identity-preserving image generation framework from ByteDance Intelligent Creation that creates highly realistic and diverse photos of individuals while maintaining their unique facial characteristics. Built on the FLUX Diffusion Transformer architecture, it allows users to generate countless new images of themselves in any imagined scenario or style by providing a reference photo and a text prompt, and it was recognized as an ICCV 2025 Highlight paper.
Key Features
Identity-preserving image generation from a single reference photo and text prompt
InfuseNet module injects identity features into the FLUX diffusion transformer via residual connections
Two model variants: aes_stage2 for aesthetics and sim_stage1 for maximum facial similarity
Multi-stage training with synthetic single-person multi-sample data for better text-image alignment
Plug-and-play compatibility with ControlNets, LoRAs and IP-Adapter
Supports optional control images for pose or structure guidance
Official ComfyUI native node and Hugging Face demo
Open-source code under Apache 2.0, models under CC BY-NC 4.0 for research use
About InfiniteYou
At its core is the InfuseNet module, a specialized neural component that injects identity features into the diffusion process through residual connections, ensuring generated images retain a strong resemblance to the original subject even as backgrounds, outfits and poses are dramatically altered. The model uses a multi-stage training strategy, including pretraining and supervised fine-tuning with synthetic single-person multi-sample data, which enhances text-image alignment and overall image aesthetics. Two model variants are offered: aes_stage2 for improved text-image alignment and visual appeal, and sim_stage1 for maximum facial similarity.
InfiniteYou is engineered with a plug-and-play design, making it compatible with a wide range of existing image generation methods, tools and extensions such as ControlNets and LoRAs, and it supports optional control images for pose or structure guidance. It is open-sourced for research and creative use, with an official ComfyUI native node and a Hugging Face demo, and it is suited to avatars, content creation, virtual try-on and personalized marketing.