Training-free Detection of Generated Videos via Spatial-Temporal Likelihoods

TL;DR

The paper introduces STALL, a training-free, likelihood-based method for detecting AI-generated videos by analyzing spatial and temporal features, outperforming existing detectors on multiple benchmarks.

Contribution

It presents STALL, a novel, training-free, probabilistic detector that effectively models spatial-temporal evidence for synthetic video detection, addressing limitations of prior methods.

Findings

STALL outperforms prior image- and video-based detectors.

It achieves state-of-the-art results on new and existing benchmarks.

The method is training-free and model-agnostic.

Abstract

Following major advances in text and image generation, the video domain has surged, producing highly realistic and controllable sequences. Along with this progress, these models also raise serious concerns about misinformation, making reliable detection of synthetic videos increasingly crucial. Image-based detectors are fundamentally limited because they operate per frame and ignore temporal dynamics, while supervised video detectors generalize poorly to unseen generators, a critical drawback given the rapid emergence of new models. These challenges motivate zero-shot approaches, which avoid synthetic data and instead score content against real-data statistics, enabling training-free, model-agnostic detection. We introduce STALL, a simple, training-free, theoretically justified detector that provides likelihood-based scoring for videos, jointly modeling spatial and temporal evidence within a probabilistic framework. We evaluate STALL on two public benchmarks and introduce ComGenVid, a new benchmark with state-of-the-art generative models. STALL consistently outperforms prior image- and video-based baselines. Code and data are available at https://omerbenhayun.github.io/stall-video.