AI Paper Atlas

Research map / Video / Video generation

Controllable First-Frame-Guided Video Editing via Mask-Aware LoRA Fine-Tuning

Chenjian Gao, Lihe Ding, Xin Cai, Zhanpeng Huang, Zibin Wang, Tianfan Xue

ICLR 2026 · Video › Video generation · approach: Diffusion-based video editing

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
use input-level mask-aware LoRA training instead of feature-level latent blending (ablation)mask-aware LoRA editing · first-frame-guided editingCLIP Score: 0.8936 -> 0.9172Improved
use input-level mask-aware LoRA training instead of feature-level latent blending (ablation)mask-aware LoRA editing · first-frame-guided editingDeQA Score: 3.5878 -> 3.8013Improved
use input-level mask-aware LoRA training instead of feature-level latent blending (ablation)mask-aware LoRA editing · first-frame-guided editingInput Similarity: 0.7402 -> 0.7608Improved
add foreground-background mask conditioning (ablation)mask-aware LoRA editing · hair color and clothing editsqualitativeImproved
add an edited later frame as appearance reference (ablation)mask-aware LoRA editing · propagated editsqualitative appearance controlImproved
use loose bounding-box or noisy mask instead of tight SAM2 mask (ablation)mask-aware LoRA editing · mask robustnessqualitativeImproved
use mask scheduling in appearance-learning stage instead of default I2V mask (ablation)mask-aware LoRA editing · appearance learningqualitativeImproved
train on 9-frame sliding windows with swap blocks instead of full 49 frames (ablation)mask-aware LoRA editing · memorypeak VRAM (MB): 21,522 -> 7,617Improved
increase blocks_to_swap from 0 to 38 (ablation)mask-aware LoRA editing · peak VRAM (9-frame segments)peak VRAM (MB): 25,097 -> 7,617Improved
vary clip length 5, 13, 21, 49 frames (ablation)mask-aware LoRA editing · CLIP Score vs frame countCLIP ScoreNo clear effect
compare with Kling1.6 on user rankingmask-aware LoRA editing · reference-guided editing user studymotion consistency rank: 1.869 -> 1.620Improved
text and rapid abstract motionmask-aware LoRA editing · failure casesqualitativeHurt

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