Inspiration
Fashion decisions happen in the real world, often with a phone already in hand: while scrolling a store, standing in a ready-to-wear shop, opening a wardrobe, or seeing an inspiring look. Smart Mirror was built as a mobile-first virtual try-on experience that makes that decision visual, personal, and fast.
What it does
Smart Mirror is a mobile-first AI virtual try-on and personal style studio. On a phone, a user creates a hybrid avatar using a face photo plus simple choices such as body shape, usual clothing size, shoe size, and skin tone. They can then see clothes, shoes, watches, glasses, and accessories on their own avatar before buying.
The product is designed for quick mobile moments:
- E-commerce: a partner store can add a See on me button under a product.
- Physical ready-to-wear shops: a shopper can photograph an item on a hanger, mannequin, or display and preview it on their avatar.
- Personal wardrobe: users can upload items they already own and receive complete look suggestions.
- Street-style inspiration: users can capture an outfit they like, with the wearer’s consent, and explore a similar look on their own avatar.
Smart Mirror includes three connected mobile spaces: Avatar Studio, AI Wardrobe, and an internal e-commerce demo. It can generate a realistic try-on image and optionally a short 5–10 second motion preview from several angles.
How we built it
The prototype is a responsive progressive web application optimized for mobile cameras, touch interaction, and mobile sharing. It uses React, a lightweight Python/Firebase backend, Three.js for interactive avatar rendering, and a generative-media pipeline. Computer vision and segmentation prepare garment and face inputs; image and video generation providers create the try-on media.
How we use Backblaze B2 and Genblaze
Backblaze B2 is the storage backbone for user face images, garment/product assets, wardrobe media, generated try-on images, videos, thumbnails, and provenance manifests.
Genblaze orchestrates the generative-media workflow. The app sends the avatar profile, selected garment, product metadata, and output choice to the workflow; Genblaze coordinates generation and returns durable media stored in B2.
Challenges
The hard part is balancing realism, privacy, and speed on a mobile connection. Smart Mirror aims to feel personal without requiring detailed measurements, while keeping user media secure and making try-on simple enough for a quick shopping decision.
What we learned
A convincing virtual try-on experience is not only about image generation. It needs mobile-friendly onboarding, reliable storage, reusable media assets, product metadata, and a flow that works for both shoppers and e-commerce partners.
What’s next
We will improve garment segmentation and fit controls, add Shopify and WooCommerce widgets, support more accessory categories, and refine 5–10 second outfit videos with front, side, and back views.
Built With
- backblaze
- python
Log in or sign up for Devpost to join the conversation.