InfiniteYou

InfiniteYou (InfU) is a robust framework for identity-preserved image generation built on Diffusion Transformers (DiTs) like FLUX. Developed by ByteDance Intelligent Creation and presented as an ICCV 2025 Highlight, InfU introduces InfuseNet—a component that injects identity features into the DiT base model via residual connections, significantly enhancing identity similarity while preserving generation quality. A multi-stage training strategy, including pretraining and supervised fine-tuning (SFT) with synthetic single-person-multiple-sample (SPMS) data, improves text-image alignment, image quality, and alleviates face copy-pasting. InfU achieves state-of-the-art performance, surpassing existing baselines such as FLUX.1-dev IP-Adapter and PuLID-FLUX. Its plug-and-play design ensures compatibility with various methods including base model replacement (e.g., FLUX.1-schnell), ControlNets, LoRAs, OminiControl (for multi-concept personalization), and IP-Adapter (for stylization). The code is released under Apache 2.0, and the model is available under CC BY-NC 4.0 for academic research.