Seeing What Matters: Generalizable AI-generated Video Detection with Forensic-Oriented Augmentation

2025-06-20

AI Summary

AI-generated video detection methods suffer from poor generalization across diverse generative models. Method: This paper proposes a forensics-oriented frequency-domain enhancement approach that leverages wavelet decomposition to localize and replace critical frequency bands, thereby guiding the detector to focus on low-level, model-agnostic artifacts introduced by generators—rather than volatile high-level semantic inconsistencies. The method employs a lightweight classifier, single-source model training, and a multi-model generalization evaluation paradigm. Contribution/Results: Trained exclusively on videos synthesized by a single generator (e.g., SVD), the method achieves significantly higher cross-model detection accuracy than state-of-the-art methods on challenging benchmarks including NOVA and FLUX. This demonstrates that the learned features are highly robust and generalize effectively across unseen generative models, without requiring multi-source training data or architectural modifications.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Generation

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Synthetic video generation is progressing very rapidly. The latest models can produce very realistic high-resolution videos that are virtually indistinguishable from real ones. Although several video forensic detectors have been recently proposed, they often exhibit poor generalization, which limits their applicability in a real-world scenario. Our key insight to overcome this issue is to guide the detector towards seeing what really matters. In fact, a well-designed forensic classifier should focus on identifying intrinsic low-level artifacts introduced by a generative architecture rather than relying on high-level semantic flaws that characterize a specific model. In this work, first, we study different generative architectures, searching and identifying discriminative features that are unbiased, robust to impairments, and shared across models. Then, we introduce a novel forensic-oriented data augmentation strategy based on the wavelet decomposition and replace specific frequency-related bands to drive the model to exploit more relevant forensic cues. Our novel training paradigm improves the generalizability of AI-generated video detectors, without the need for complex algorithms and large datasets that include multiple synthetic generators. To evaluate our approach, we train the detector using data from a single generative model and test it against videos produced by a wide range of other models. Despite its simplicity, our method achieves a significant accuracy improvement over state-of-the-art detectors and obtains excellent results even on very recent generative models, such as NOVA and FLUX. Code and data will be made publicly available.

Problem

• Detect AI-generated videos with poor generalization in existing methods
• Identify intrinsic low-level artifacts instead of high-level semantic flaws
• Improve detector generalizability using forensic-oriented augmentation and wavelet decomposition

Innovation

• Focuses on low-level artifacts for detection
• Uses wavelet-based forensic data augmentation
• Improves generalization with single-model training