Audio visual method detects edits in long videos with sparse changes
AV-FABLE: Audio-Visual Manipulation Forensics in Long-Form Videos
Computer Vision and Pattern RecognitionMultimedia
Summary
It can be hard to find small edited parts in long videos that mostly look real. The authors created a big test set called AV-FABLE with many long videos that have small audio or visual edits, to help find these tricky changes. They tried using big audio-visual language models to spot edits but got almost random results. However, when they used parts of these models in a new way that looks separately at audio and video, they could find the edited parts much better. Their approach works well even on new videos not seen before.
What this means in practice
- •For video security teams: Identify small fake audio or visual edits within long surveillance or media videos to improve forensic detection accuracy.
- •For video editing software developers: Integrate specialized detection tools that highlight brief and sparse manipulations in lengthy videos from separate audio and visual perspectives.
Authors
Yuesheng Huang, Hoi Fan Au, Jiaqi Hou, Ruicheng Zhang, Jiarui He, Haocheng Yang, Bin Hu, Kewei Zhang, Weijian Deng, Jun Zhou, Xiu Li
Abstract
Localized AI manipulations in long-form videos often affect brief events within otherwise authentic content, making video-level classification insufficient for forensic analysis. Existing audio-visual deepfake benchmarks largely focus on short talking-head clips, while long-form temporal forensics remains predominantly visual. We introduce AV-FABLE, a dual-timeline benchmark for sparse manipulations in open-scene long-form audio-visual videos. AV-FABLE contains 16,650 pristine--manipulated twin pairs (998 hours) created with 27 editing configurations, with separate audio and visual timelines that enable modality-specific temporal localization and expose errors hidden by joint-timeline evaluation. On AV-FABLE, zero-shot prompting of audio-visual large language models (AV-LLMs) yields near-chance detection and poor temporal localization. In contrast, supervised readouts of their frozen perception features recover strong forensic signals, suggesting manipulation evidence is encoded in these representations but not effectively exposed through language outputs. Motivated by this observation, we propose NFSR(Native Forensics via Separated Readouts), which reads frozen AV-LLM features with separate audio and visual temporal decoders. NFSR achieves the best audio-visual localization performance in-domain and across all three out-of-distribution settings, including transfer to held-out videos from an external long-video forensic benchmark. AV-FABLE, code, and model weights will be publicly released.