Instruction-guided video editing requires executing a requested change while preserving the source video's appearance and temporal structure. We present SemantiScene-Edit, a video editor that couples a multimodal encoder with a pretrained video diffusion transformer. The encoder jointly processes source frames and the...
Xin-Yu Chen, Song-Han Wu, Di Shao et al.· 0 citations
VAR has gained widespread popularity due to its next-scale prediction paradigm. However, it faces substantial performance bottlenecks when handling complex scenes with multiple objects and attributes. Existing diffusion-based enhancement methods fail to adequately address the unique challenge of cross-scale error propa...
Zhen-Nan Chen, Tianxing Shi, Pengcheng Xu et al.· 0 citations
VideoX-Qwen, an integrated data-construction and model-training framework for general instruction-based video editing, and a unified Qwen-Wan editor that combines multimodal semantic conditioning with dense source-video latent guidance provide a practical foundation for more capable instruction-driven video editing.
JJiahang Li, Di Shao, Xin-Yu Chen et al.· arXiv.org· 0 citations
The Self-Corrective Multi-Agent Refinement Framework is proposed, which decouples physics-aware error correction from code generation through specialized agents and nearly triples the physical-assertion pass rate of the best proprietary baseline and retains its advantage under cross-engine transfer.
Tianyidan Xie, Pei-Yu Wang, Jia-Xin Hu et al.· arXiv.org· 7 citations· ⚡2
Linguistic Trajectory Encoding (LTE) is introduced, which compresses dynamic object motion histories via a hybrid representation combining natural language descriptions, sparse spatial anchors, and visual anchors, and adapts compression to motion complexity by anchoring periods without reliable observations to the last...
Tianyidan Xie, Shen-Yi Wang, Qiang Tang et al.· 0 citations
InstructVVT is proposed, an instruction-driven and reference-guided video virtual try-on framework based on a Diffusion Transformer that operates without inference-time spatial priors that outperforms state-of-the-art open-source methods in garment fidelity, structural preservation, and temporal consistency, despite re...
Di Shao, Song-Han Wu, Xin-Yu Chen et al.· 0 citations
Evaluation of state-of-the-art open-source and closed-source MLLMs reveals that current MLLMs function as lossy summarizers rather than faithful memorizers, highlighting the need for architectures with genuine long-term spatiotemporal memory.