Research map / Video / Video generation
One-Shot Free-View Neural Talking-Head Synthesis for Video Conferencing
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 3D keypoint decomposition and synthesis model instead of FOMM | TalkingHead-1KH face reconstruction · TalkingHead-1KH | FID: 55.35 -> 52.08 | Helped |
| use the proposed 3D keypoint decomposition and synthesis model instead of FOMM | VoxCeleb2 face reconstruction · VoxCeleb2 | L1: 12.66 -> 10.74 | Helped |
| use the proposed model instead of fs-vid2vid | cross-identity motion transfer · VoxCeleb2 | CSIM: 0.593 -> 0.753 | Helped |
| use the proposed model instead of fs-vid2vid | cross-identity motion transfer · TalkingHead-1KH | FID: 52.72 -> 46.99 | Helped |
| use the proposed model for face frontalization instead of pSp | face frontalization · face frontalization | Identity (%): 57.3 -> 94.3 | Helped |
| use the proposed model for face frontalization instead of RaR | face frontalization · face frontalization | FID: 78.81 -> 23.87 | Helped |
| replace the decomposed canonical-keypoint, pose, and expression pipeline with direct keypoint prediction (ablation) | proposed talking-head model · TalkingHead-1KH face reconstruction | FID: 52.08 -> 58.55 | Hurt |
| replace 3D keypoint warping with 2D keypoint warping (ablation) | proposed talking-head model · TalkingHead-1KH face reconstruction | FID: 52.08 -> 58.75 | Hurt |
| reduce the number of keypoints from 20 to 10 (ablation) | proposed talking-head model · TalkingHead-1KH face reconstruction | FID: 52.08 -> 56.27 | Hurt |
| reduce the number of keypoints from 20 to 15 (ablation) | proposed talking-head model · TalkingHead-1KH face reconstruction | FID: 52.08 -> 54.36 | Hurt |
| increase FOMM filter size to double its model size | FOMM · face reconstruction | performance | Hurt |
| use adaptive keypoint dropout to reduce the transmitted keypoints (ablation) | video conferencing compression · talking-head video conferencing | keypoints sent: 20 -> 11.52 | Helped |
Closest papers by meaning
- FaceTalk: Audio-Driven Motion Diffusion for Neural Parametric Head Models (CVPR 2024 · similarity 0.86)
- SyncTalk: The Devil is in the Synchronization for Talking Head Synthesis (CVPR 2024 · similarity 0.85)
- From Faces to Voices: Learning Hierarchical Representations for High-quality Video-to-Speech (CVPR 2025 · similarity 0.85)
- MoDiTalker: Motion-Disentangled Diffusion Model for High-Fidelity Talking Head Generation (AAAI 2025 · similarity 0.85)
- Synergizing Motion and Appearance: Multi-Scale Compensatory Codebooks for Talking Head Video Generation (CVPR 2025 · similarity 0.85)
- Occlusion-Insensitive Talking Head Video Generation via Facelet Compensation (AAAI 2025 · similarity 0.84)
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.