Features
Maskless virtual try-on without segmentation masks
Pixel-space generation on raw RGB pixels
~5 second inference on H100 GPUs
972M parameter MMDiT architecture
Pose keypoint conditioning via grayscale images
Category embedding support (tops, bottoms, one-pieces)
Open-source under Apache 2.0 license
Synthetic triplet training for maskless inference
Fast generation mode in app and API
Up to 4K image generation and upscaling
AI video generation up to 1080p
Product-to-Model, Model Swap, Model Creation tools
Assistant with conversation history and generation settings
Team collaboration with role-based access
API for custom integrations
Customizable AI models (size, emotion, pose, skin tone, hair)
Enterprise AI model library for consistent branding
Stylezone collaboration for real-time asset sharing
E-commerce-ready content generation without physical samples
Validated digital twins with accurate fit consistency
IP and copyright ownership of AI model assets
Integration with 3D design tools (VStitcher, Lotta)
Wholesale presentation with on-brand AI models
Localized, diverse visuals for global campaigns
Cost reduction on photoshoots and licensing fees
Faster time-to-market via reduced sampling
Design validation and review in Stylezone
