Kolors Virtual Try-On: AI Clothing Try-On by Kuaishou 167

Kolors Virtual Try-On is an AI try-on technology developed by the Kwai Kolors team. Through Kolors AI Try-On, users can upload their own photos and see themselves wearing different outfits within seconds. Based on the Kolors foundation model, Kolors AI Try-On introduces a clothing SKU preservation network to extract and represent apparel detail features. The goal is to let users instantly see how clothing looks on them, easily mix and match different styles, enable buyers to make purchasing decisions more efficiently, and reduce return costs for sellers.

Kolors Virtual Try-On Kolors AI Try-On Technology Features:

Preserves clothing style details: Supports multiple clothing types and styles, including tops, long dresses, etc., while preserving details such as patterns and text designs, achieving a consistent SKU-level apparel replication effect.

Natural try-on results on people: Kolors AI Try-On can generate natural try-on effects that fit the person, as if they were photographed wearing the clothes themselves. It supports various scenarios including indoor portraits and outdoor street shots, achieving a transition from flat-laid clothing to being worn on the body, with generated results that conform to physical objective laws.

End-to-end asset generation capabilities: The Kolors large model supports generating model images and backgrounds across different ages, genders, and ethnicities, meeting e-commerce model asset generation needs both domestically and internationally, as well as the full workflow from model asset images to model short videos.

Kolors Virtual Try-On Kolors AI Try-On Application Scenarios:

• Buyer try-on: Helps buyers see the actual effect of clothing before purchasing, improving the efficiency of purchase decisions.

• E-commerce asset generation: Provides sellers with the ability to quickly update apparel images, short videos, and other e-commerce assets, shortening the iteration and update time for new product listings.

• Outfit-change effects: Provides "outfit change" effects in short video content creation, producing fun and useful everyday outfit content for sharing.

Kolors Virtual Try-On Kolors AI Try-On Technical Principles:

Kolors Virtual Try-On technology introduces a clothing SKU preservation network to extract and represent apparel detail features. At the same time, it uses person pose and Mask images as conditional features to achieve clothing-changing effects while preserving the person's pose. In addition, it reuses Kolors' background inpainting technology and is trained on high-quality person-clothing paired data to improve try-on results.

The main goal of Kolors Virtual Try-On Kolors AI Try-On is to help users try different clothing styles without the hassle of physical shopping. It aims to enhance the online shopping experience by reducing issues related to size and color mismatches.

Kolors Virtual Try-On Kolors AI Try-On Access:

Free trial of Kolors AI Try-On: https://huggingface.co/spaces/Kwai-Kolors/Kolors-Virtual-Try-On

Kolors open-source project on GitHub: https://github.com/Kwai-Kolors/Kolors

Kolors model on Hugging Face: https://huggingface.co/Kwai-Kolors/Kolors

Kling AI platform: https://klingai.kuaishou.com/text-to-image/new