ByteDance, owner of TikTok has released a new AI model called InfiniteYou, designed to create personalized versions of a person’s image while keeping their identity intact. The model, known as InfU, uses advanced image generation techniques to let users change the look or context of a photo—without losing the person’s core features.
InfiniteYou was developed to fix common problems in AI-generated portraits, like poor image quality or faces that don’t resemble the original person. It does this with a new method called InfuseNet, which injects identity data into the image creation process. The result is better-looking photos that still look like the user.
Developers can also customize the model further. InfiniteYou supports upgrades using ControlNets and LoRAs—popular tools that help fine-tune how AI generates images. It’s compatible with different base models, including FLUX.1-dev and FLUX.1-schnell, which allow faster or more flexible results.
ByteDance has made the code available on GitHub, and a public demo is live on Hugging Face for anyone who wants to try it.
Link: https://bytedance.github.io/InfiniteYou/
