Research map / Video / Video generation
Stable Video Portraits
Mirela Ostrek, Justus Thies
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 |
|---|---|---|---|
| reduce training data from 100% to 6.25% of frames (ablation) | Stable Video Portraits · Obama sequence | PSNR: 33.78 -> 20.99 | Hurt |
| reduce training data from 100% to 50% of frames (ablation) | Stable Video Portraits · Obama sequence | LPIPS: 0.016 -> 0.016 | No clear effect |
| use SVP instead of PECHead on perceptual metrics | Stable Video Portraits · Monocular head avatars | LPIPS: 0.034 -> 0.026 | Improved |
| use SVP instead of PECHead on FID | Stable Video Portraits · Monocular head avatars | FID: 23 -> 11 | Improved |
| SVP versus best baselines on pixel metrics (PSNR, SSIM, MSE) | Stable Video Portraits · Monocular head avatars | PSNR/SSIM/MSE: PSNR 31.4 vs 31.4; SSIM 0.9445 vs 0.9487; MSE 0.0010 vs 0.0009 | No clear effect |
| drop the previous-frame information in the denoising (original scheme) (ablation) | Stable Video Portraits · Video generation | temporal stability (qualitative) | Hurt |
| increase influence of previous frame (w_p) (ablation) | Stable Video Portraits denoising · Video generation | smoothness vs sharpness | Mixed |
| use 5 consecutive 3DMM meshes as conditioning instead of 1 (ablation) | Stable Video Portraits ControlNet · Video generation | temporal smoothness (qualitative) | Improved |
| use person-specific fine-tuning instead of a person-generic model trained on FFHQ and CelebA-HQ (ablation) | Stable Video Portraits · Video generation | identity consistency (qualitative) | Improved |
| lower ControlNet strength below 0.5 for face morphing (ablation) | Celeb face morphing · Face morphing | temporal stability and 3DMM control | Hurt |
Closest papers by meaning
- IM-Portrait: Learning 3D-aware Video Diffusion for Photorealistic Talking Heads from Monocular VideosC (CVPR 2025 · similarity 0.90)
- Morphable Diffusion: 3D-Consistent Diffusion for Single-image Avatar Creation (CVPR 2024 · similarity 0.90)
- Real3D-Portrait: One-shot Realistic 3D Talking Portrait Synthesis (ICLR 2024 · similarity 0.89)
- CAP4D: Creating Animatable 4D Portrait Avatars with Morphable Multi-View Diffusion Models (CVPR 2025 · similarity 0.89)
- DAWN: Dynamic Frame Avatar with Non-autoregressive Diffusion Framework for Talking head Video Generation (ICLR 2025 · similarity 0.88)
- Human-3Diffusion: Realistic Avatar Creation via Explicit 3D Consistent Diffusion Models (NeurIPS 2024 · similarity 0.87)
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.
