AI Paper Atlas

Research map / Video / Video generation

TGT: Text-Grounded Trajectories for Locally Controlled Video Generation

Guofeng Zhang, Angtian Wang, Jacob Zhiyuan Fang, Liming Jiang, Haotian Yang, Bo Liu, Yiding Yang, Guang Chen, Longyin Wen, Alan Yuille, Chongyang Ma

CVPR 2026 · Video › Video generation · approach: Compositional text-to-video diffusion · 2 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
add sparse-track tuning after dense-track training (ablation)LACA · DAVISEPE: 58.01 -> 45.28Improved
add sparse-track tuning after dense-track training (ablation)LACA · DAVISCLIP-T (local): 0.2394 -> 0.2447Improved
add sparse-track tuning after dense-track training (ablation)LACA · DAVISCLIP-T (global): 0.3307 -> 0.3312No clear effect
add Gaussian masking to sparse-track LACA (ablation)LACA · DAVISEPE: 45.28 -> 25.11Improved
add Gaussian masking to sparse-track LACA (ablation)LACA · DAVISCLIP-T (local): 0.2447 -> 0.2527Improved
use dual-CFG instead of global-only guidance (ablation)TGT · DAVISEPE: 91.38 -> 25.11Improved
use dual-CFG instead of local-only guidance (ablation)TGT · DAVISCLIP-T (global): 0.3117 -> 0.3314Improved
use dual-CFG instead of local-only guidance (ablation)TGT · DAVISEPE: 43.29 -> 25.11Improved
use dual-CFG instead of combined global-with-local guidance (ablation)TGT · DAVISEPE: 53.30 -> 25.11Improved
use dual-CFG instead of combined global-with-local guidance (ablation)TGT · DAVISCLIP-T (local): 0.2491 -> 0.2527Improved
add extended global+local prompts describing motion and location to WanWan2.2 14B · DAVISEPE: 265.03 -> 180.36Improved
use TGT instead of Tora trajectory controlWan2.1 14B · DAVISEPE: 47.41 -> 25.11Improved

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