How it works
Dress Your Pet
Dress Your Pet is a creative AI image generation demo that brings imagination to life. Pick a pet, mix and match a wardrobe of up to five clothing items, and watch the model generate a brand-new, photorealistic image of your pet wearing the chosen pieces as adorable, pet-sized outfits.
Business applications
1
Virtual Try-On for E-Commerce
This underlying generative composition technology can revolutionize online apparel. Customers can upload a photo of themselves and seamlessly "try on" multiple clothing items or accessories at once to see a preview of how an outfit looks before purchasing, significantly boosting conversion rates and reducing returns.
2
Custom Merchandise & Visual Marketing
Brands can create highly engaging, interactive storefronts or ad campaigns where users dynamically mix-and-match product features to generate customized, highly shareable visual assets.
High-level technical workflow
1
Asset selection & retrieval
The user selects a pet and up to five clothing items. The backend loads all corresponding image files from disk and prepares them for AI input.
2
Visual input composition
All selected clothing images are composited into a single labeled grid ("Item 1: Top Hat", "Item 2: Bow Tie") using server-side image processing, so the model receives exactly two clean visual inputs (the pet photo and the clothing grid) alongside a structured text prompt.
3
Multimodal generation request
The composed inputs are sent to the image generation model. The text prompt explicitly references both images by position and names each clothing item, giving the model visual and linguistic grounding for every piece.
A system prompt acts as a persistent creative contract enforcing that the animal stays an animal, specifying anatomical placement rules for each clothing type, and directing the model toward funny and charming rather than uncanny or human-like results.
5
Image extraction & delivery
The used model returns the generated image as raw bytes in the response. The backend extracts it, base64-encodes it, and sends it to the frontend for immediate display and download — no intermediate storage required.