ResearchBenchmarksAudio-driven 3D Talking Head Generation on VASA generated synthetic video 1.0 (test)Audio-driven 3D Talking Head Generation on VASA generated synthetic video 1.0 (test)Follow8.121ScVASA-3D5.155965.925736.69557.46527Dec 16, 2025Evaluation resultsMethodMethodLinksScSDID SimilarityUS: Video QualityUS: Overall PreferenceVASA-3D2025.128.1216.930.78744.2993.91VASA-3DTraining Iterations=20kTraining Iterations=20k2025.127.987.002———TalkingGaussian2025.126.7018.10610.79712.380.72GeneFace2025.125.9229.60660.78571.730.72ER-NeRF2025.125.9218.77880.77321.821.08MimicTalk2025.125.2710.93680.77482.233.58