AI Paper Atlas

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

ICCV 2025 · Video › Video generation · approach: Compositional text-to-video diffusion · 11 citations (OpenAlex)

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 changedWhereMeasuredEffect
train camera first, then IDs, then depth (instead of depth, camera, IDs) (ablation)FullDiT · FullBench camera-to-videoTransErr: 6.57 -> 3.31Improved
train camera first, then IDs, then depth (instead of IDs, camera, depth) (ablation)FullDiT · FullBench camera-to-videoTransErr: 7.43 -> 3.31Improved
train camera first, then IDs, then depth (instead of IDs, camera, depth) (ablation)FullDiT · FullBench identities-to-videoDINO-I: 42.71 -> 46.22Improved
train camera first, then IDs, then depth (instead of depth first) (ablation)FullDiT · FullBench depth-to-videoMAE: 14.76 -> 14.71No clear effect
use three progressive training stages instead of one joint stage (ablation)FullDiT · FullBench camera-to-videoRotErr: 2.69 -> 1.20Improved
use three progressive training stages instead of one joint stage (ablation)FullDiT · FullBench depth-to-videoMAE: 32.88 -> 14.71Improved
isolate camera as its own first stage (two-stage instead of one-stage) (ablation)FullDiT · FullBench camera-to-videoTransErr: 6.19 -> 4.14Improved
add a third stage (Camera+ID+Depth) after Camera+ID (ablation)FullDiT · FullBench camera-to-videoTransErr: 4.20 -> 3.31Improved
use full-attention condition fusion instead of an adapter (ablation)FullDiT · FullBench camera-to-videoCamMC: 4.17 -> 3.98Improved
use full-attention condition fusion instead of an adapter (ablation)FullDiT · FullBench camera-to-videoDynamic: 28.42 -> 30.53Improved
use full-attention condition fusion instead of an adapter (ablation)FullDiT · FullBench camera-to-videoSmoothness: 96.41 -> 96.40No clear effect
full attention vs ConceptMaster on identity injection with the same backbone and dataFullDiT · FullBench identities-to-videoDINO-I: 39.97 -> 46.22Improved

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.

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.

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

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy