Anti-Prompt: Image Protection against Text-Guided Image-to-Video Generation

2026-07-01

AI Summary

This work addresses the risks of copyright infringement and privacy leakage in text-guided image-to-video (I2V) generation by exploiting the model’s reliance on textual cues. The authors propose a novel defense mechanism that injects imperceptible adversarial perturbations into input images to attenuate the influence of text guidance while reinforcing purely visual pathways. This intervention induces structural distortions and temporal inconsistencies during video synthesis, effectively safeguarding source images. The approach integrates adversarial perturbation generation with text-conditioned denoising modulation and introduces an interpretable evaluation protocol based on Video-LLMs. Experiments demonstrate that the method achieves strong, generalizable protection across two state-of-the-art I2V models without compromising visual fidelity under normal viewing conditions.

Abstract

Recent advances in Image-to-Video generation allow a single image to be animated into a convincing video under text guidance, raising serious copyright and privacy risks. We propose Anti-Prompt, an image protection approach that injects imperceptible perturbations into an image, inducing visible inconsistencies and structural failures in text-guided I2V generation. Our method is motivated by a simple empirical observation. When text guidance is removed from modern I2V models, generation quality degrades markedly, not only in motion realism but also in subject preservation, structural coherence, and temporal consistency. Building on this insight, Anti-Prompt exploits the model reliance on textual guidance by attenuating text-conditioned interactions during denoising while strengthening visual-only pathways. To further systematically evaluate protection effectiveness, we introduce a Video-LLM-assisted evaluation protocol that provides interpretable, frame-grounded analyses of generation artifacts and inconsistencies. Experiments on two representative I2V architectures demonstrate that our method achieves strong protection performance while improving efficiency and cross-model transferability.

Problem

• image protection
• text-guided image-to-video generation
• copyright risk
• privacy risk
• generation inconsistency

Innovation

• Anti-Prompt
• image protection
• text-guided image-to-video generation
• adversarial perturbation
• Video-LLM evaluation