Research map / Video / Video generation
Latent-Reframe: Enabling Camera Control for Video Diffusion Models without Training
Zhenghong Zhou, Jie An, Jiebo Luo
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 time-aware point clouds instead of a single static global point cloud (ablation) | Latent-Reframe · qualitative video comparison | motion preservation and image quality | Improved |
| apply latent reframing at diffusion step 8 instead of step 0 (ablation) | Latent-Reframe · qualitative video comparison | visual artifacts | Improved |
| apply latent reframing at diffusion step 8 instead of step 16 (ablation) | Latent-Reframe · qualitative video comparison | point cloud accuracy and content consistency | Improved |
| reduce noise of known regions by 3 steps instead of 0 steps (ablation) | Latent-Reframe · qualitative video comparison | artifacts | Improved |
| reduce noise of known regions by 3 steps instead of 5 steps (ablation) | Latent-Reframe · qualitative video comparison | blurriness | Improved |
| use MonST3R instead of DUSt3R for point cloud extraction (ablation) | Latent-Reframe · qualitative video comparison | detail preservation | Improved |
| use MonST3R instead of MASt3R for point cloud extraction (ablation) | Latent-Reframe · qualitative video comparison | pose accuracy | Improved |
| training-free Latent-Reframe versus fine-tuned CameraCtrl | AnimateDiff · RealEstate10K poses, 10 prompts | FID: 92.40 -> 60.18 | Improved |
| training-free Latent-Reframe versus fine-tuned MotionCtrl | AnimateDiff · RealEstate10K poses, 10 prompts (463 successful samples) | rotation error: 1.23 -> 2.29 | Hurt |
Closest papers by meaning
- Training-free Camera Control for Video Generation (ICLR 2025 · similarity 0.90)
- VD3D: Taming Large Video Diffusion Transformers for 3D Camera Control (ICLR 2025 · similarity 0.89)
- Efficient Video Diffusion Models via Content-Frame Motion-Latent Decomposition (ICLR 2024 · similarity 0.89)
- AC3D: Analyzing and Improving 3D Camera Control in Video Diffusion Transformers (CVPR 2025 · similarity 0.89)
- Generative Inbetweening: Adapting Image-to-Video Models for Keyframe Interpolation (ICLR 2025 · similarity 0.88)
- DiffPerformer: Iterative Learning of Consistent Latent Guidance for Diffusion-based Human Video Generation (CVPR 2024 · 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.
