Research map / Video / Video generation
FullDiT: Video Generative Foundation Models with Multimodal Control via Full Attention
Xuan Ju, Weicai Ye, Quande Liu, Qiulin Wang, Xintao Wang, Pengfei Wan, Di Zhang, Kun Gai, Qiang Xu
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 |
|---|---|---|---|
| train camera first, then IDs, then depth (instead of depth, camera, IDs) (ablation) | FullDiT · FullBench camera-to-video | TransErr: 6.57 -> 3.31 | Improved |
| train camera first, then IDs, then depth (instead of IDs, camera, depth) (ablation) | FullDiT · FullBench camera-to-video | TransErr: 7.43 -> 3.31 | Improved |
| train camera first, then IDs, then depth (instead of IDs, camera, depth) (ablation) | FullDiT · FullBench identities-to-video | DINO-I: 42.71 -> 46.22 | Improved |
| train camera first, then IDs, then depth (instead of depth first) (ablation) | FullDiT · FullBench depth-to-video | MAE: 14.76 -> 14.71 | No clear effect |
| use three progressive training stages instead of one joint stage (ablation) | FullDiT · FullBench camera-to-video | RotErr: 2.69 -> 1.20 | Improved |
| use three progressive training stages instead of one joint stage (ablation) | FullDiT · FullBench depth-to-video | MAE: 32.88 -> 14.71 | Improved |
| isolate camera as its own first stage (two-stage instead of one-stage) (ablation) | FullDiT · FullBench camera-to-video | TransErr: 6.19 -> 4.14 | Improved |
| add a third stage (Camera+ID+Depth) after Camera+ID (ablation) | FullDiT · FullBench camera-to-video | TransErr: 4.20 -> 3.31 | Improved |
| use full-attention condition fusion instead of an adapter (ablation) | FullDiT · FullBench camera-to-video | CamMC: 4.17 -> 3.98 | Improved |
| use full-attention condition fusion instead of an adapter (ablation) | FullDiT · FullBench camera-to-video | Dynamic: 28.42 -> 30.53 | Improved |
| use full-attention condition fusion instead of an adapter (ablation) | FullDiT · FullBench camera-to-video | Smoothness: 96.41 -> 96.40 | No clear effect |
| full attention vs ConceptMaster on identity injection with the same backbone and data | FullDiT · FullBench identities-to-video | DINO-I: 39.97 -> 46.22 | Improved |
Later papers that compared with this method
1 later accepted paper reports full-dit as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- UnityVideo: Unified Multi-Modal Multi-Task Learning for Enhancing World-Aware Video Generation (CVPR 2026 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- 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
- DiTCtrl: Exploring Attention Control in Multi-Modal Diffusion Transformer for Tuning-Free Multi-Prompt Longer Video Generation (CVPR 2025 · similarity 0.87)
- VSTAR: Generative Temporal Nursing for Longer Dynamic Video Synthesis (ICLR 2025 · similarity 0.87)
- DuoGen: Towards Autonomous Interleaved Multimodal Generation (CVPR 2026 · similarity 0.87)
- ConditionVideo: Training-Free Condition-Guided Video Generation (AAAI 2024 · similarity 0.86)
- Video-Bench: Human-Aligned Video Generation Benchmark (CVPR 2025 · similarity 0.86)
- ShotAdapter: Text-to-Multi-Shot Video Generation with Diffusion Models (CVPR 2025 · similarity 0.86)
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.
