Paper 2605.15824
FashionChameleon: Towards Real-Time and Interactive Human-Garment Video Customization
- Published
- May 2026
- Research lab
- Independent
- Citations
- 3
- GitHub
- 256 stars
01 In brief
Summary
FashionChameleon is a real-time and interactive framework for human-garment video customization, enabling users to switch garments during generation while preserving motion coherence.
It uses three key techniques: a Teacher Model with In-Context Learning trained on single-garment data to implicitly handle garment switching; Streaming Distillation with In-Context Learning for efficient few-step generation and consistent long-video extrapolation; and Training-Free KV Cache Rescheduling for interactive multi-garment switching.
The model achieves 23.8 FPS on a single H200 GPU, 30-180x faster than baselines, and outperforms them on garment consistency and temporal smoothness in the HGC-Bench benchmark.
It also supports long-video extrapolation and interactive customization, with a data curation pipeline yielding 62K training triplets.
02 From the paper
Abstract
Human-centric video customization, particularly at the garment level, has shown significant commercial value. However, existing approaches cannot support low-latency and interactive garment control, which is crucial for applications such as e-commerce and content creation. This paper studies how to achieve interactive multi-garment video customization while preserving motion coherence using only single-garment video data. We present FashionChameleon, a real-time and interactive framework for human-garment customization in autoregressive video generation, where users can interactively switch garment during generation. FashionChameleon consists of three key techniques: (i) Instead of training on multi-garment video data, we train a Teacher Model with In-Context Learning on a single reference-garment pair. By retaining the image-to-video training paradigm while enforcing a mismatch between the reference and garment image, the model is encouraged to implicitly preserve coherence during single-garment switching. (ii) To achieve consistency and efficiency during generation, we introduce Streaming Distillation with In-Context Learning, which fine-tunes the model with in-context teacher forcing and improves extrapolation consistency via gradient-reweighted distribution matching distillation. (iii) To extend the model for interactive multi-garment video customization, we propose Training-Free KV Cache Rescheduling, which includes garment KV refresh, historical KV withdraw, and reference KV disentangle to achieve garment switching while preserving motion coherence. Our FashionChameleon uniquely supports interactive customization and consistent long-video extrapolation, while achieving real-time generation at 23.8 FPS on a single GPU, 30-180$\times$ faster than existing baselines.