Research map / Video / Video generation
I2V3D: Controllable Image-to-video Generation with 3D Guidance
Zhiyuan Zhang, Dongdong Chen, Jing Liao
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 |
|---|---|---|---|
| customize LoRA with multi-view augmentation images of the foreground instead of the single input image (ablation) | I2V3D keyframe generation · qualitative ablation | view generalization | Improved |
| add extended attention across keyframes (ablation) | I2V3D keyframe generation · qualitative ablation | keyframe consistency | Improved |
| combine depth ControlNet and rendered feature control in interpolation (ablation) | I2V3D 3D-guided interpolation · qualitative ablation | alignment with rendered video | Improved |
| remove depth control (rendered features only) (ablation) | I2V3D 3D-guided interpolation · qualitative ablation | geometry accuracy | Hurt |
| remove rendered feature control (depth only) (ablation) | I2V3D 3D-guided interpolation · qualitative ablation | fine detail alignment | Hurt |
| use 3D-guided pipeline instead of skeleton, trajectory or box control baselines | I2V3D · Nonhuman-like objects | SSIM: 0.570 (ISculpting) -> 0.757 | Improved |
| use 3D-guided pipeline versus ISculpting | I2V3D · Human-like characters | CLIP-I: 0.916 -> 0.901 | No clear effect |
| reduce 3D guidance strength from 1.0 to 0.5 and 0.2 (ablation) | I2V3D · supplementary videos | background dynamism | 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.
- AnimateAnyone — Animate Anyone: Consistent and Controllable Image-to-Video Synthesis for Character Animation (CVPR 2024) · better on 4 of 4 shared results
- DragAnything — Drag Anything: Motion Control for Anything using Entity Representation (ECCV 2024) · better on 4 of 4 shared results
- MotionBooth — MotionBooth: Motion-Aware Customized Text-to-Video Generation (NeurIPS 2024) · better on 3 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
- Generative Rendering: Controllable 4D-Guided Video Generation with 2D Diffusion Models (CVPR 2024 · similarity 0.90)
- GEN3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control (CVPR 2025 · similarity 0.88)
- 3DTrajMaster: Mastering 3D Trajectory for Multi-Entity Motion in Video Generation (ICLR 2025 · similarity 0.88)
- Interactive3D: Create What You Want by Interactive 3D Generation (CVPR 2024 · similarity 0.88)
- SG-I2V: Self-Guided Trajectory Control in Image-to-Video Generation (ICLR 2025 · similarity 0.87)
- Extrapolating and Decoupling Image-to-Video Generation Models: Motion Modeling is Easier Than You Think (CVPR 2025 · 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.
