FAME: Fairness-aware Attention-modulated Video Editing

TL;DR

FAME is a novel video editing method that reduces gender bias in generated videos by integrating fairness-aware attention mechanisms, ensuring coherent and unbiased results without sacrificing prompt accuracy.

Contribution

FAME introduces a fairness-aware attention modulation framework that mitigates gender stereotypes in video editing while maintaining temporal consistency and prompt alignment.

Findings

FAME achieves superior fairness and semantic fidelity on the FairVE benchmark.

It effectively reduces gender bias in profession-related video prompts.

FAME maintains temporal coherence and prompt accuracy in edited videos.

Abstract

Training-free video editing (VE) models tend to fall back on gender stereotypes when rendering profession-related prompts. We propose \textbf{FAME} for \textit{Fairness-aware Attention-modulated Video Editing} that mitigates profession-related gender biases while preserving prompt alignment and temporal consistency for coherent VE. We derive fairness embeddings from existing minority representations by softly injecting debiasing tokens into the text encoder. Simultaneously, FAME integrates fairness modulation into both temporal self attention and prompt-to-region cross attention to mitigate the motion corruption and temporal inconsistency caused by directly introducing fairness cues. For temporal self attention, FAME introduces a region constrained attention mask combined with time decay weighting, which enhances intra-region coherence while suppressing irrelevant inter-region interactions. For cross attention, it reweights tokens to region matching scores by incorporating fairness sensitive similarity masks derived from debiasing prompt embeddings. Together, these modulations keep fairness-sensitive semantics tied to the right visual regions and prevent temporal drift across frames. Extensive experiments on new VE fairness-oriented benchmark \textit{FairVE} demonstrate that FAME achieves stronger fairness alignment and semantic fidelity, surpassing existing VE baselines.