Research map / Video / Video generation
Anti-Prompt: Image Protection against Text-Guided Image-to-Video Generation
Yeonghwan Song, Chanhui Lee, Jinsoo Park, Jeany Son
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add text suppression loss L_sup to encoder attack (ablation) | Anti-Prompt · CogVideoX white-box | Video-LLM average (lower is stronger): 4.11 -> 3.92 | Improved |
| Add text suppression loss L_sup to encoder attack (ablation) | Anti-Prompt · CogVideoX white-box VBench | Subject Consistency: 91.96 -> 92.94 | Mixed |
| Add visual dominance loss L_vis to encoder attack (ablation) | Anti-Prompt · CogVideoX white-box | Video-LLM average (lower is stronger): 4.11 -> 3.59 | Improved |
| Combine L_sup and L_vis (ablation) | Anti-Prompt · CogVideoX white-box | Video-LLM average (lower is stronger): 3.59 -> 3.25 | Improved |
| Use Anti-Prompt instead of I2VGuard (white-box) | Anti-Prompt · CogVideoX VBench | I2V Subject: 94.58 -> 91.20 | Improved |
| Use Anti-Prompt instead of I2VGuard (white-box) | Anti-Prompt · LTX-Video Video-LLM average | average: 3.28 -> 3.12 | Improved |
| Use Anti-Prompt under unseen prompts | Anti-Prompt · CogVideoX Video-LLM | average: 3.86 -> 2.99 | Improved |
| Transfer perturbation across architectures (LTX to CogVideoX) | Anti-Prompt · black-box | Video-LLM average: 4.47 -> 4.15 | Improved |
| Transfer from CogVideoX to LTX-Video | Anti-Prompt · black-box VBench | I2V Subject: 94.52 vs 94.61 | Mixed |
| Use losses on intermediate attention statistics instead of generating reference video | Anti-Prompt · CogVideoX-5B efficiency | total PFLOPs: 54.25 -> 26.82 | Improved |
| Compare perturbation imperceptibility with I2VGuard | Anti-Prompt · LTX-Video | DISTS: 0.130 -> 0.138 | Hurt |
| Compare perturbation imperceptibility with I2VGuard | Anti-Prompt · CogVideoX | PSNR: 32.86 -> 35.10 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- I2VGuard — I2VGuard: Safeguarding Images against Misuse in Diffusion-based Image-to-Video Models (CVPR 2025) · better on 12 of 13 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- I2VGuard: Safeguarding Images against Misuse in Diffusion-based Image-to-Video Models (CVPR 2025 · similarity 0.88)
- SAFREE: Training-Free and Adaptive Guard for Safe Text-to-Image And Video Generation (ICLR 2025 · similarity 0.87)
- OVERT: A Benchmark for Over-Refusal Evaluation on Text-to-Image Models (NeurIPS 2025 · similarity 0.86)
- T2VSafetyBench: Evaluating the Safety of Text-to-Video Generative Models (NeurIPS 2024 · similarity 0.86)
- UPAM: Unified Prompt Attack in Text-to-Image Generation Models Against Both Textual Filters and Visual Checkers (ICML 2024 · similarity 0.86)
- Anti-Diffusion: Preventing Abuse of Modifications of Diffusion-Based Models (AAAI 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
