Research map / 3D vision / Face and shape
SyncTalk++: High-Fidelity and Efficient Synchronized Talking Heads Synthesis Using Gaussian Splatting
Ziqiao Peng, Wentao Hu, Jiao Ma, Xiangyu Zhu, X. M. Zhang, Hao Zhao, Hui Tian, Jun He, Hongyan Liu, Zhaoxin Fan
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add Portrait-Sync Generator (ablation) | SyncTalk++ · Head reconstruction (Portrait mode) | PSNR: 36.3779 -> 39.5748 | Improved |
| add Portrait-Sync Generator (ablation) | SyncTalk++ · Component ablation (May) | PSNR: 33.759 -> 39.093 | Improved |
| use the audio-visual encoder instead of HuBERT features (ablation) | SyncTalk++ · Component ablation (May) | LMD: 3.250 -> 2.715 | Improved |
| use the audio-visual encoder instead of HuBERT features (ablation) | SyncTalk++ · Component ablation (May) | PSNR: 37.638 -> 39.093 | Improved |
| use 3D blendshape Facial Animation Capture instead of AU units (ablation) | SyncTalk++ · Component ablation (May) | PSNR: 37.264 -> 39.093 | Improved |
| add Head-Sync Stabilizer (ablation) | SyncTalk++ · Component ablation (May) | PSNR: 29.193 -> 39.093 | Improved |
| add Semantic Weighting to the head tracker (ablation) | SyncTalk++ · Component ablation (May) | PSNR: 34.526 -> 39.093 | Improved |
| add Torso Restorer for OOD audio (ablation) | SyncTalk++ · OOD audio torso quality | BRISQUE: 33.2916 -> 25.4147 | Improved |
| add Torso Restorer for OOD audio (ablation) | SyncTalk++ · OOD audio torso quality | NIQE: 15.2476 -> 14.3012 | Improved |
| add Torso Restorer for OOD audio (ablation) | SyncTalk++ · OOD audio torso quality | HyperIQA: 62.0307 -> 66.1958 | Improved |
| initialize all Gaussian attributes (s, r, SH, alpha) instead of only SH and alpha (ablation) | canonical 3D head · Initialization of canonical 3D head | PSNR: 33.759 -> 31.468 | Hurt |
| use random initialization instead of SH, alpha initialization (ablation) | canonical 3D head · Initialization of canonical 3D head | PSNR: 33.759 -> 29.902 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- SyncTalk — SyncTalk: The Devil is in the Synchronization for Talking Head Synthesis (CVPR 2024) · better on 4 of 5 shared results
- TalkingGaussian — TalkingGaussian: Structure-Persistent 3D Talking Head Synthesis via Gaussian Splatting (ECCV 2024) · better on 2 of 3 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
- SyncTalk: The Devil is in the Synchronization for Talking Head Synthesis (CVPR 2024 · similarity 0.95)
- GoHD: Gaze-oriented and Highly Disentangled Portrait Animation with Rhythmic Poses and Realistic Expressions (AAAI 2025 · similarity 0.90)
- PointTalk: Audio-Driven Dynamic Lip Point Cloud for 3D Gaussian-based Talking Head Synthesis (AAAI 2025 · similarity 0.90)
- Perceptually Accurate 3D Talking Head Generation: New Definitions, Speech-Mesh Representation, and Evaluation Metrics (CVPR 2025 · similarity 0.89)
- FaceTalk: Audio-Driven Motion Diffusion for Neural Parametric Head Models (CVPR 2024 · similarity 0.89)
- EmotiveTalk: Expressive Talking Head Generation through Audio Information Decoupling and Emotional Video Diffusion (CVPR 2025 · similarity 0.88)
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.
