FlowVVTON: Flow-Guided Mask-Free Video Virtual Try-On
2026-08-31 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors developed FlowVVTON, a new way to put clothes on people in videos without needing special masks or body landmarks. Instead, their method uses optical flow during training to help the computer understand how the person is moving between frames, which makes the clothes fit more smoothly over time. They train their model in two steps to first learn how to align clothes and body shapes and then improve motion consistency. Tests show their approach works better than previous methods, especially for keeping the outfit looking stable across frames.
video virtual try-onoptical flowtemporal consistencymask-freemotion supervisionlatent lossspatial alignmentvideo frame synthesisattention mechanismsflow-warping
Authors
Shengyao Chen, Xianbing Sun, Liqing Zhang, Jianfu Zhang
Abstract
Video virtual try-on aims to transfer a target garment onto a moving person across video frames. Current methods rely on human parsing masks or pose keypoints that frequently fail under large motions and occlusions, causing boundary artifacts and temporal inconsistency. A further limitation is that most approaches rely solely on attention mechanisms for temporal modeling, providing no explicit motion supervision. We propose FlowVVTON, a mask-free framework that eliminates parsing mask dependency entirely. Optical flow is used solely as a training-time supervision signal: a flow-warped latent loss, applied across all layers of the generation model, enforces multi-scale temporal consistency by aligning adjacent-frame features under explicit physical motion constraints. A two-stage training strategy establishes mask-free spatial alignment before introducing flow-guided temporal supervision. Experiments on TikTokDress show that FlowVVTON outperforms baselines by substantial margins, particularly in temporal consistency (5.7$\times$ VFID-R improvement over SwiftTry), while requiring no segmentation masks, pose keypoints, or region annotations at any stage.