AI Paper Atlas

Research map / Video / Video generation

EDTalk: Efficient Disentanglement for Emotional Talking Head Synthesis

Shuai Tan, Bin Ji, Mengxiao Bi, ye pan

ECCV 2024 · Video › Video generation · approach: Talking face generation · 62 citations (OpenAlex)

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 orthogonality constraint on bases (ablation)EDTalk · MEADAcc_emo: 38.71 -> 67.32Improved
add base banks storing visual priors (ablation)EDTalk · MEADAcc_emo: 9.677 -> 67.32Improved
add Emotion Enhancement Module (ablation)EDTalk · MEADAcc_emo: 49.367 -> 67.32Improved
add feature-level loss L_fea (ablation)EDTalk · MEADAcc_emo: 54.34 -> 67.32Improved
add self-reconstruction loss L_self (ablation)EDTalk · MEADAcc_emo: 44.23 -> 67.32Improved
add sync loss L^m_sync (ablation)EDTalk audio-to-lip · MEADSync_conf: 4.007 -> 8.115Improved
add expression disentanglement to lip and pose spaces (ablation)EDTalk · MEADSync_conf: 8.474 -> 8.115 Sync_conf; 16.75 -> 67.32 Acc_emoMixed
add reconstruction loss L^m_rec (ablation)EDTalk audio-to-lip · MEADPSNR: 21.955 -> 21.628 PSNR; 5.447 -> 8.115 Sync_confMixed
increase mouth bank size from 5 to 20 bases (ablation)EDTalk · MEADAcc_emo: 63.53 -> 67.32 (40 bases: 63.12)Improved
increase expression bank size beyond 10 bases (ablation)EDTalk · MEADAcc_emo: 67.32 -> 61.34 (20), 59.65 (40)Hurt
train disentanglement only lightweight modules on small dataEDTalk · training costtraining time: about one hour on 2 GPUs for mouth-pose vs over 2 days for DPEImproved

Later papers that compared with this method

11 later accepted papers report EDTalk as a baseline in their results tables; 10 of them report a better number on most of the results they share with it.

3 more not listed.

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