🎨 MagicMakeup: High-Precision Makeup Transfer Using DiT
MagicMakeup has been introduced—a framework based on Diffusion Transformer (DiT) for transferring makeup from a reference to a face without losing identity. The model is built on FLUX.1-Kontext-dev and utilizes TARG and CMPG modules for precise regional control and the separation of style from facial features.
🌍 MagicMakeup solves the problems of inaccurate positioning and facial feature distortion during editing. The use of the DiT architecture sets a new standard for specialized i2i tasks in the digital beauty industry.
👤 The tool allows for virtually "trying on" any makeup while maintaining full recognizability, which is useful for content creators and developers of beauty applications.
Source 1: https://vivocameraresearch.github.io/magicmakeup/ Source 2: https://github.com/vivoCameraResearch/Magic-Makeup