AI Paper Atlas

Research map / Video / Video generation

Ariadne's Thread of LipSync: Unraveling Forgeries via Inconsistency between Lip Motions and Head Poses

Tianyi She, Jiawei Liu, Weifeng Liu, Hanqing Zhao, Weiming Zhang, Kejiang Chen

ICML 2026 · Video › Video generation · approach: Talking face generation

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add contrastive alignment loss between lip and pose embeddings (ablation)LipDA Stage I · AVLipsACC: 80.56 -> 97.90Improved
Add head pose branch (ablation)LipDA Stage I · AVLipsACC: 74.78 -> 97.90Improved
Add lip branch (ablation)LipDA Stage I · AVLipsACC: 53.50 -> 97.90Improved
Replace ResNet-18 lip encoder with CLIP-ViT (ablation)LipDA · AVLipsACC: 97.90 -> 85.73Hurt
Replace ResNet-18 lip encoder with MobileNet (ablation)LipDA · AVLipsACC: 97.90 -> 96.23Hurt
Use flatten or pooling pose features instead of LSTM (ablation)LipDA · AVLipsACC: 97.90 -> 95.27 (Flatten); 97.90 -> 97.20 (Pooling)Hurt
Add audio MSM score to visual-only LipDA (ablation)LipDA · TalkHeadBenchAUC: 98.65 -> 97.50Mixed
Add audio MSM score to visual-only LipDA (ablation)LipDA · LipSync-AAUC: 97.83 -> 99.42Improved
Increase sliding window length T excessively (ablation)LipDA · detection accuracy vs window TaccuracyHurt
Increase total video length up to about 8 seconds (ablation)LipDA · detection accuracy vs video lengthaccuracyImproved
Use landmark-based pose features under Gaussian noise perturbation (ablation)LipDA · AVLips Level-5 Gaussian noiseAUC: 99.72 -> 90.98Hurt
Use LipDA under Level-3 visual corruptionsLipDA · AVLipsAUC (average): 85.53 -> 99.11 (vs LipForensics)Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy