MagicID: Hybrid Preference Optimization for ID-Consistent and Dynamic-Preserved Video Customization
Hengjia Li, Lifan Jiang, Xi Xiao, Tianyang Wang, Hongwei Yi, Boxi Wu, Deng Cai
2025-03-21
Summary
This paper is about creating videos that can change a person's appearance while keeping their identity consistent and the movements realistic.
What's the problem?
It's hard to make videos where you can change someone's look because the person's identity can get messed up, and the movements can look stiff or unnatural.
What's the solution?
The researchers created a new system called MagicID that uses a special way of learning from videos to make sure the person's identity stays consistent and the movements look realistic.
Why it matters?
This work matters because it can be used to create more realistic and personalized videos for entertainment, communication, and other applications.
Abstract
Video identity customization seeks to produce high-fidelity videos that maintain consistent identity and exhibit significant dynamics based on users' reference images. However, existing approaches face two key challenges: identity degradation over extended video length and reduced dynamics during training, primarily due to their reliance on traditional self-reconstruction training with static images. To address these issues, we introduce MagicID, a novel framework designed to directly promote the generation of identity-consistent and dynamically rich videos tailored to user preferences. Specifically, we propose constructing pairwise preference video data with explicit identity and dynamic rewards for preference learning, instead of sticking to the traditional self-reconstruction. To address the constraints of customized preference data, we introduce a hybrid sampling strategy. This approach first prioritizes identity preservation by leveraging static videos derived from reference images, then enhances dynamic motion quality in the generated videos using a Frontier-based sampling method. By utilizing these hybrid preference pairs, we optimize the model to align with the reward differences between pairs of customized preferences. Extensive experiments show that MagicID successfully achieves consistent identity and natural dynamics, surpassing existing methods across various metrics.