Outfit Anyone

Outfit Anyone is a virtual try-on research project from the Institute for Intelligent Computing at Alibaba Group. Its project page presents a conditional diffusion model that takes an image of a person, an image of a garment and a text prompt, then generates an image of the person wearing that clothing. The model uses two separate streams to process the person and the garment before fusing them, which the authors say helps handle garment deformation and keep details consistent. The system has two parts: a zero-shot try-on network that produces the initial image and a post-hoc refiner that enhances clothing and skin texture.

The page shows results for full outfits and single garments, unusual fashion items, different body shapes and anime characters, and a demonstration combining Outfit Anyone with the Animate Anyone pose-to-video model to produce moving characters in new outfits. It links to the research team's GitHub and Hugging Face pages.

Outfit Anyone is presented as research rather than a commercial product. The page states that the project is intended solely for academic research and effect demonstration and that no commercial benefits are derived from it, and that most models and clothing images come from the internet and public datasets such as VITON and DressCode. No pricing, service terms or assured public access were found on the reviewed page, and no try-on was run for this listing.

Virtual try-on with real people's photos raises consent and likeness concerns, so anyone experimenting with related demos should only use images they have the right to use.