Leveraging Pre-Trained Visual Models for AI-Generated Video Detection

2025-07-17

AI Summary

Existing AI-generated video detection methods are largely confined to deepfake face manipulation and struggle to generalize to increasingly realistic text-to-video (T2V) models. Method: We propose a zero-shot/low-shot detection framework leveraging large-scale pretrained vision models (e.g., CLIP), requiring no model fine-tuning. It extracts frame-level visual features, applies temporal pooling, and employs a lightweight linear classifier to capture intrinsic discriminative signals between real and synthetic videos. Contribution/Results: Evaluated on VID-AID—a newly constructed, large-scale benchmark comprising over 10,000 videos spanning nine state-of-the-art T2V models—the framework achieves >90% average detection accuracy. It significantly transcends face-centric detection paradigms, demonstrating strong generalization across diverse generative architectures and prompting modalities. This work provides a scalable, model-agnostic technical foundation for misinformation mitigation, privacy preservation, and multimedia content security.

Technology Category

• Computer Vision: Large Vision Models
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Recent advances in Generative AI (GenAI) have led to significant improvements in the quality of generated visual content. As AI-generated visual content becomes increasingly indistinguishable from real content, the challenge of detecting the generated content becomes critical in combating misinformation, ensuring privacy, and preventing security threats. Although there has been substantial progress in detecting AI-generated images, current methods for video detection are largely focused on deepfakes, which primarily involve human faces. However, the field of video generation has advanced beyond DeepFakes, creating an urgent need for methods capable of detecting AI-generated videos with generic content. To address this gap, we propose a novel approach that leverages pre-trained visual models to distinguish between real and generated videos. The features extracted from these pre-trained models, which have been trained on extensive real visual content, contain inherent signals that can help distinguish real from generated videos. Using these extracted features, we achieve high detection performance without requiring additional model training, and we further improve performance by training a simple linear classification layer on top of the extracted features. We validated our method on a dataset we compiled (VID-AID), which includes around 10,000 AI-generated videos produced by 9 different text-to-video models, along with 4,000 real videos, totaling over 7 hours of video content. Our evaluation shows that our approach achieves high detection accuracy, above 90% on average, underscoring its effectiveness. Upon acceptance, we plan to publicly release the code, the pre-trained models, and our dataset to support ongoing research in this critical area.

Problem

• Detect AI-generated videos beyond deepfakes
• Distinguish real from generated generic video content
• Combat misinformation using pre-trained visual models

Innovation

• Leverages pre-trained visual models
• Extracts features for detection
• Uses linear classification layer