Original title: BooM-VVT: Boosting Mask-Free Video Virtual Try-On with Image-Level Pseudo Data
Wei Zhang, Xin Li, Peishu Shi, Jialin Gao +3 · September 2026
Video virtual try-on systems aim to show a person wearing new clothes in a realistic video. Older methods rely on digital masks to outline where clothes go, but big movements or blocked views break them. Making fake video data to train mask-free models is also too expensive. To solve this, researchers built BooM-VVT, a mask-free try-on framework based on keyframes, which are selected important video frames. It trains in stages using cheaper fake single-image data instead of full videos to learn clothing locations. It also uses Garment-Sensitive Keyframe Sampling to pick keyframes focusing on clothing-related body areas. A tool named Frame-Shared 3D-RoPEA technique for aligning spatial and temporal features between frames. links space and time between keyframes and target video frames to transfer fine details. Finally, they built OmniView, a large multi-view dataset for tricky camera angles and diverse try-on tasks. Tests show BooM-VVT beats existing methods in clothing detail fidelity and video consistency across time.
Imagine you're packing for a trip and need to decide what clothes to bring, but you only have a few photos of your wardrobe instead of seeing everything laid out. You might miss an outfit that doesn't match or forget a key piece. In this paper, the system uses images instead of full videos to learn how clothes fit, like using photos to plan an outfit without seeing the whole closet.
Video virtual try-on helps people see how clothes look on them online, but current methods are unreliable when people move a lot or are partially hidden. This work makes the technology more realistic and cheaper to develop, which could improve online shopping and content creation.
The abstract does not mention specific limitations, but it implies that previous methods had issues with motion and occlusions. The work focuses on improving consistency and fidelity, but details about computational cost or real-world application are not provided.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
BooM-VVT: Boosting Mask-Free Video Virtual Try-On with Image-Level Pseudo Data — Wei Zhang, Xin Li, Peishu Shi et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.