Self-Supervised Video Forensics by Audio-Visual Anomaly Detection

Chao Feng, Ziyang Chen, Andrew Owens
2/14/2026

No official code found. Best implementation: cfeng16/audio-visual-forensics (113 stars, MIT, last commit 2024-05-12).

Abstract

Manipulated videos often contain subtle inconsistencies between their visual and audio signals. We propose a video forensics method, based on anomaly detection, that can identify these inconsistencies, and that can be trained solely using real, unlabeled data. We train an autoregressive model to generate sequences of audio-visual features, using feature sets that capture the temporal synchronization between video frames and sound. At test time, we then flag videos that the model assigns low probability. Despite being trained entirely on real videos, our model obtains strong performance on the task of detecting manipulated speech videos. Project site: https://cfeng16.github.io/audio-visual-forensics.

View on arXivView PDF

Code Implementations(1)

cfeng16/audio-visual-forensicsCommunity100%
MIT

[CVPR 2023 (Highlight)] Self-Supervised Video Forensics by Audio-Visual Anomaly Detection

1138PythonDec 1, 20222 years agoMIT

Similar Papers

AVadCLIP: Audio-Visual Collaboration for Robust Video Anomaly Detection

87% match2025

AVAR-Net: A Lightweight Audio-Visual Anomaly Recognition Framework with a Benchmark Dataset

86% match2025

MULDE: Multiscale Log-Density Estimation via Denoising Score Matching for Video Anomaly Detection

86% match1 repo542026

Multimodal Attention-Enhanced Feature Fusion-based Weekly Supervised Anomaly Violence Detection

86% match2024

Detecting AI-Generated Video via Frame Consistency

86% match1 repo152024

Harnessing Large Language Models for Training-Free Video Anomaly Detection

86% match1 repo1492026

Cite this paper

@article{feng2026selfsupervised,
  title  = {Self-Supervised Video Forensics by Audio-Visual Anomaly Detection},
  author = {Chao Feng and Ziyang Chen and Andrew Owens},
  year   = {2026},
  eprint = {2301.01767},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/pdf/2301.01767},
  journal = {CVPR 2023 2023}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2301.01767.svg)](https://codeofpaper.com/paper/2301.01767)

Discussion