MagicID：融合偏好優化的身份一致性與動態保持影片定制技術

MagicID: Hybrid Preference Optimization for ID-Consistent and Dynamic-Preserved Video Customization

March 16, 2025

作者: Hengjia Li, Lifan Jiang, Xi Xiao, Tianyang Wang, Hongwei Yi, Boxi Wu, Deng Cai

cs.AI

摘要

視頻身份定製旨在基於用戶的參考圖像，生成高保真度且保持身份一致性的視頻，同時展現出顯著的動態效果。然而，現有方法面臨兩個主要挑戰：視頻長度增加時的身份退化，以及訓練過程中動態效果的減弱，這主要歸因於它們依賴於傳統的靜態圖像自重建訓練。為解決這些問題，我們引入了MagicID，這是一個新穎的框架，旨在直接促進生成符合用戶偏好且身份一致、動態豐富的視頻。具體而言，我們提出構建具有明確身份和動態獎勵的成對偏好視頻數據，用於偏好學習，而非固守傳統的自重建方法。為應對定製偏好數據的限制，我們引入了一種混合採樣策略。該策略首先通過利用從參考圖像衍生的靜態視頻來優先保持身份，然後使用基於前沿的採樣方法提升生成視頻中的動態運動質量。通過利用這些混合偏好對，我們優化模型以對齊定製偏好對之間的獎勵差異。大量實驗表明，MagicID成功實現了身份一致性和自然動態效果，在多項指標上超越了現有方法。

English

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.