One-step face swapping method matches multi-step quality rapidly
ALADDIN: Adversarial Latent Diffusion Distillation for One-Step Identity Swapping
Computer Vision and Pattern Recognition
Summary
Face swapping means putting one person's face onto another's photo while keeping things like pose and lighting the same. The authors created ALADDIN, a new method that can do this very quickly in just one step but still keep the swapped face realistic and natural looking. It works better than older fast methods and almost as well as slower methods that take many steps. They also found that how methods are usually tested is a bit unfair and fixed that to better compare their results.
What this means in practice
- •For mobile app developers: Create real-time face swapping features that deliver high-quality results efficiently on smartphones.$Commercial implications: Enables fast, high-quality face swap filters for social media or entertainment apps, improving user experience and reducing computational cost.
- •For visual effects artists: Generate realistic identity swaps in video production workflows with fewer computational resources and faster turnaround.
Authors
Ivan Shchekotov, Timofei Gritsaev, Viacheslav Pirogov, Dmitry Vetrov
Abstract
Face swapping transfers a source identity onto a target image while preserving the target's pose, expression, hair, and lighting. The most realistic current methods are diffusion-based, but they need many network evaluations per image. GAN-based methods are fast, yet trained from scratch they must learn realism from the adversarial signal alone and produce visibly weaker results. We present ALADDIN (Adversarial Latent Diffusion Distillation for One-Step Identity Swapping), a one-step face swapper obtained by initialising the generator from a pre-trained multi-step flow-matching model and training it adversarially under identity and target-preservation constraints. We expose a controllable trade-off between identity transfer and target-attribute preservation by training a student model with triplet pseudo-label supervision. We additionally identify a bias in the standard evaluation protocol, which favours methods trained against the same identity encoder used to score them, and re-evaluate all baselines with an independent one. ALADDIN is state-of-the-art among one-step methods in realism and target preservation at comparable identity transfer, and matches the best multi-step model with 28 times fewer network evaluations and an order of magnitude fewer training pairs.