APPLE: Attribute-Preserving Pseudo-Labeling for Diffusion-Based Face Swapping

TL;DR

APPLE tackles the challenge of preserving target attributes during diffusion-based face swapping in the absence of ground-truth swapped data. It introduces a teacher–student framework where a teacher is trained with a conditional deblurring objective and attribute-aware inversion to produce high-quality pseudo-triplets to supervise a student that learns via direct image editing. Key contributions include reformulating face swapping as conditional deblurring, introducing attribute-aware inversion, and demonstrating state-of-the-art attribute preservation with competitive identity transfer on FFHQ while requiring no external conditioning at inference. The approach yields more photorealistic, target-faithful outputs and offers a practical path for real-world deployment.

Abstract

Face swapping aims to transfer the identity of a source face onto a target face while preserving target-specific attributes such as pose, expression, lighting, skin tone, and makeup. However, since real ground truth for face swapping is unavailable, achieving both accurate identity transfer and high-quality attribute preservation remains challenging. In addition, recent diffusion-based approaches attempt to improve visual fidelity through conditional inpainting on masked target images, but the masked condition removes crucial appearance cues of target, resulting in plausible yet misaligned attributes. To address these limitations, we propose APPLE (Attribute-Preserving Pseudo-Labeling), a diffusion-based teacher-student framework that enhances attribute fidelity through attribute-aware pseudo-label supervision. We reformulate face swapping as a conditional deblurring task to more faithfully preserve target-specific attributes such as lighting, skin tone, and makeup. In addition, we introduce an attribute-aware inversion scheme to further improve detailed attribute preservation. Through an elaborate attribute-preserving design for teacher learning, APPLE produces high-quality pseudo triplets that explicitly provide the student with direct face-swapping supervision. Overall, APPLE achieves state-of-the-art performance in terms of attribute preservation and identity transfer, producing more photorealistic and target-faithful results.

Figures (15)

• Figure 1: APPLE (Attribute-Preserving Pseudo-Labeling) successfully transfers the identity of a source (top left) onto a target (bottom left) while accurately preserving target attributes (e.g. pose, expression, skin tone, lighting) across ethnicity, input variations, and gender.
• Figure 2: Comparison of conditioning methods. Compared to conditional inpainting widely used in existing works zhao2023diffswapyu2025refacehan2024faceadapter, proposed conditional deblurring strategy achieves largely improved attribute (e.g. lighting) preservation of targets.
• Figure 3: Overall architecture of the proposed method. We propose APPLE, a diffusion-based teacher-student framework that focuses on improving attribute-preservation. (a) To improve target-attribute preservation, we propose training a teacher with conditional deblurring rather than conditional inpainting widely used in existing works zhao2023diffswapyu2025refacehan2024faceadapter. (b) When constructing pseudo-triplet with teacher, we propose attribute-aware inversion which further improves attribute preservation in inference time. Note that inversion noise cannot be used during training due to its non-Gaussian property. (c) Student is trained with pseudo-triplet generated by teacher. Thanks to pseudo-supervision, student even surpass the teacher, achieving state-of-the-art attribute preservation while maintaining high identity similarity.
• Figure 4: Comparison of conditioning configuration for inversion. (Top) Visualization of target-inverted noise and random gaussian noise via PCA. When attribute-condition is used, inverted noise encodes more semantic information compared to the others. (Bottom) Results of face swapping when each inverted noise used. Using attribute-only conditioned noise yields most make-up preserved results without introducing artifacts.
• Figure 5: Qualitative results on FFHQ. Compared to existing baselines, APPLE effectively preserves the target image’s attributes while faithfully transferring the source identity.