Research map / 3D vision / Face and shape
GaussianSpeech: Audio-Driven Personalized 3D Gaussian Avatars
Shivangi Aneja, Artem Sevastopolsky, Tobias Kirschstein, Justus Thies, Angela Dai, Matthias Nießner
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 |
|---|---|---|---|
| use the proposed avatar initialization instead of GaussianAvatars | 3DGS head avatar · Avatar reconstruction (Subject 4) | PSNR / # Gaussians: 26.53 -> 29.90 PSNR; 98083 -> 32379 Gaussians | Improved |
| add perceptual losses (ArcFace global + VGG patch) (ablation) | avatar initialization · Avatar reconstruction (Subject 4) | PSNR: 27.03 -> 29.90 | Improved |
| add wrinkle regularization (ablation) | avatar initialization · Avatar reconstruction (Subject 4) | PSNR: 28.10 -> 29.90 | Improved |
| add mouth region subdivision (ablation) | avatar initialization · Avatar reconstruction (Subject 4) | PSNR: 28.35 -> 29.90 | Improved |
| add expression-dependent color MLP (ablation) | avatar initialization · Avatar reconstruction (Subject 4) | PSNR: 28.93 -> 29.90 | Improved |
| add alignment mask in audio cross-attention (ablation) | GaussianSpeech sequence model · Audio-driven animation (Subject 4) | LSE-D: 12.66 -> 11.15 | Improved |
| personalize lip encoder via audio fine-tuning (ablation) | GaussianSpeech sequence model · Audio-driven animation (Subject 4) | LSE-D: 11.78 -> 11.15 | Improved |
| add wrinkle features for expression prediction (ablation) | GaussianSpeech sequence model · Audio-driven animation (Subject 4) | PSNR: 23.14 -> 24.97 | Improved |
| fine-tune color MLP and Gaussian latents with predicted expressions (ablation) | GaussianSpeech · Audio-driven animation (Subject 4) | PSNR: 23.96 -> 24.97 | Improved |
| GaussianSpeech vs best mesh-based baseline combined with GaussianAvatars | GaussianSpeech · Audio-driven animation (frontal camera) | LSE-D: 11.61 -> 11.25 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- GaussianAvatar — GaussianAvatar: Towards Realistic Human Avatar Modeling from a Single Video via Animatable 3D Gaussians (CVPR 2024) · better on 3 of 3 shared results
- SyncTalk — SyncTalk: The Devil is in the Synchronization for Talking Head Synthesis (CVPR 2024) · better on 1 of 1 shared result
- RAD-NeRF — Rad-NeRF: Ray-decoupled Training of Neural Radiance Field (NeurIPS 2024) · better on 1 of 1 shared result
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
- GaussianAvatars: Photorealistic Head Avatars with Rigged 3D Gaussians (CVPR 2024 · similarity 0.89)
- Gaussian Eigen Models for Human Heads (CVPR 2025 · similarity 0.89)
- Gaussian Head Avatar: Ultra High-fidelity Head Avatar via Dynamic Gaussians (CVPR 2024 · similarity 0.89)
- GASP: Gaussian Avatars with Synthetic Priors (CVPR 2025 · similarity 0.89)
- 3D Gaussian Head Avatars with Expressive Dynamic Appearances by Compact Tensorial Representations (CVPR 2025 · similarity 0.89)
- Human Gaussian Splatting: Real-time Rendering of Animatable Avatars (CVPR 2024 · similarity 0.89)
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.
