Paper Detail
Xiao Cao, Yansong Qu, Xiangzhen, Chang, Wen Xiao, Jiakui Hu, Heyuan Li, Jialun Liu, Zhiyong Huang, Xuelong Li
Mask-free video object insertion has emerged as a challenging task, requiring harmonious integration of reference objects into source videos. However, existing methods struggle when references exhibit severe stylistic domain gaps with the source scene. To overcome this, we propose \textit{\textbf{Smart-Insertion-V}}, an end-to-end \textbf{Dual-Stream} framework that concurrently conducts video insertion and image style transfer. Within this framework, the image stream synchronously guides the video generation process, while a \textbf{Closed-loop Feedback} mechanism is further incorporated to ensure robust insertion. Inevitably, integrating these diverse conditioning signals results in feature entanglement and style leakage. To tackle this issue, we design \textbf{Dual-World-View RoPE} to distinguish different signals via spatial-temporal offsets without incurring heavy training overhead. Furthermore, to facilitate spatial grounding and stylistic adaptation, we introduce a \textbf{Decoupled Guidance Module} that leverages a Vision-Language Model for semantic reasoning while preserving original temporal guidance with native text encoder. To bridge data gap for harmonious reference insertion task, we propose a data curation pipeline and will release an \textbf{open-source dataset}. Experiments demonstrate that our method can insert objects into plausible positions while achieving the most harmonious results.
No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.
No ranking explanation is available yet.
No tags.
@article{cao2026smart,
title = {Smart-Insertion-V: Photorealistic Video Insertion via a Closed-Loop Feedback Dual-Stream Framework},
author = {Xiao Cao and Yansong Qu and Xiangzhen and Chang and Wen Xiao and Jiakui Hu and Heyuan Li and Jialun Liu and Zhiyong Huang and Xuelong Li},
year = {2026},
abstract = {Mask-free video object insertion has emerged as a challenging task, requiring harmonious integration of reference objects into source videos. However, existing methods struggle when references exhibit severe stylistic domain gaps with the source scene. To overcome this, we propose \textbackslash{}textit\{\textbackslash{}textbf\{Smart-Insertion-V\}\}, an end-to-end \textbackslash{}textbf\{Dual-Stream\} framework that concurrently conducts video insertion and image style transfer. Within this framework, the image stream synchronously guides the vi},
url = {https://arxiv.org/abs/2605.23891},
keywords = {cs.CV},
eprint = {2605.23891},
archiveprefix = {arXiv},
}
{}