FLUX-Makeup: Identity-consistent makeup transfer(paper and comfyUI)

FLUX-Makeup introduces a groundbreaking framework for makeup transfer that emphasizes high fidelity, identity consistency, and robustness, eliminating the need for auxiliary face-control components. This innovative approach enables users to apply makeup using a simple 'source + reference' input method, which enhances user experience by offering a more natural interaction. The significance of this development lies in its potential impact on personal grooming technologies and virtual makeup applications, catering to both consumers and the beauty industry.

The underlying architecture features a unique decoupled feature injection mechanism facilitated by RefLoRAInjector, which accurately extracts makeup-related information while preventing issues like identity collapse and background distortion. The team also developed the HQMT dataset, a high-quality paired makeup collection comprising over 50,000 samples, to enhance model training. By providing pretrained weights for easy implementation, along with an inference code, FLUX-Makeup empowers developers and researchers within the AI/ML community to explore robust makeup transfer solutions, pushing the boundaries of what is possible in the realm of digital cosmetics.