Research map / Video / Video generation
OmniVDiff: Omni Controllable Video Diffusion for Generation and Understanding
Dianbing Xi, Jiepeng Wang, Yuanzhi Liang, Xi Qiu, Yuchi Huo, Rui Wang, Chi Zhang, Xuelong Li
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 |
|---|---|---|---|
| add modality embedding for generation vs conditioning roles (ablation) | OmniVDiff · VBench text-to-video | weighted average: 71.54 -> 72.78 | Improved |
| add adaptive modality control strategy (AMCS) (ablation) | OmniVDiff · VBench text-to-video | weighted average: 71.21 -> 72.78 | Improved |
| add modality-specific projection heads (MSPH) (ablation) | OmniVDiff · VBench text-to-video | weighted average: 71.35 -> 72.78 | Improved |
| joint multi-modal training vs rgb-only CogVideoX | CogVideoX1.5-5B · VBench text-to-video | weighted average: 72.25 -> 72.78 | Improved |
| joint multi-modal training vs rgb-only CogVideoX | CogVideoX1.5-5B · VBench text-to-video | dynamic degree: 53.98 -> 49.69 | Hurt |
| add 10k synthetic ground-truth depth samples to training (ablation) | OmniVDiff · ScanNet zero-shot video depth | AbsRel: 0.125 -> 0.100 | Improved |
| train with pseudo labels from VDA-S instead of expert directly | OmniVDiff · ScanNet zero-shot video depth | AbsRel: VDA-S 0.110 vs OmniVDiff 0.125 | Hurt |
| unified model vs expert segmentation model | OmniVDiff · COCO Val 2017 point-based segmentation | 1-IoU (Max): 57.0 -> 56.0 | No clear effect |
| add two extra modalities (segmentation, depth) to output vs rgb-only CogVideoX | OmniVDiff · inference time | time (s): 41s -> 44s | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DepthCrafter — DepthCrafter: Generating Consistent Long Depth Sequences for Open-world Videos (CVPR 2025) · better on 1 of 1 shared result
- ControlVideo — ControlVideo: Training-free Controllable Text-to-video Generation (ICLR 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
- BIVDiff: A Training-Free Framework for General-Purpose Video Synthesis via Bridging Image and Video Diffusion Models (CVPR 2024 · similarity 0.90)
- VD3D: Taming Large Video Diffusion Transformers for 3D Camera Control (ICLR 2025 · similarity 0.88)
- DiTCtrl: Exploring Attention Control in Multi-Modal Diffusion Transformer for Tuning-Free Multi-Prompt Longer Video Generation (CVPR 2025 · similarity 0.87)
- OSV: One Step is Enough for High-Quality Image to Video Generation (CVPR 2025 · similarity 0.87)
- Generative Omnimatte: Learning to Decompose Video into Layers (CVPR 2025 · similarity 0.87)
- FreeLong: Training-Free Long Video Generation with SpectralBlend Temporal Attention (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.
