• 

• 

• 

• 

• 

• 

Calliope

Inspiration

Calliope started with a simple idea: “Tinder for outfits.” Put two looks in front of someone, let them keep a favorite, and give it a new challenger. I wanted to explore whether a series of small, easy decisions could make choosing a complete outfit feel less overwhelming.

The opportunity was to make every comparison inform the next, so shoppers could discover their preferences as they went. A favorite should emerge through choices and feedback, with the occasion, budget, and existing wardrobe still respected. That became my goal for Calliope: a stylist that builds on what you like until you find an outfit you want to wear—and can actually buy.

What it does

Calliope is a personal shopping stylist that turns your choices and feedback into progressively better-matched outfits you can actually buy, while respecting your budget and practical needs.

Brief your stylist. Share your photo, occasion, budget, and preferences. You can also add a garment you want to build an outfit around. A suggested color palette provides a starting direction for the first looks.

Find your favorites. Compare complete outfits on your own photo, keeping a favorite as new challengers appear. Your keeps and passes help guide subsequent suggestions. Calliope also asks what you liked, so your explicit feedback takes priority over its interpretation.

Make it yours. Ask for changes to individual pieces without rebuilding the outfit. Share the look whenever you want a second opinion.

Bring it home. Review the selected pieces, choose sizes, and check prices before opening the actual products in the merchant’s cart. Garments you already own remain part of the look without becoming new purchases.

How YouCam powers Calliope

YouCam gives Calliope a starting direction—and shoppers a way to judge what comes next.

What Calliope needs	YouCam capability	What it makes possible
Give the stylist a warm start.	AI Facial Color Tones Analyzer	Calliope interprets color observations into a suggested palette to guide the first outfits.
Remove the mental work of picturing an outfit.	Virtual try-on	Shoppers compare complete looks on their own photo. Their choices help shape the next suggestions.

YouCam makes visualization part of the feedback loop: shoppers see, choose, and steer what Calliope recommends next.

How we built it

Build on real products. Calliope imports the merchant catalog through MCP and classifies supported garments into a searchable index. Selected catalog garments become the inputs to YouCam’s virtual try-on, keeping each preview connected to the products the shopper can purchase.

Let preferences guide constrained selection. Keeps and passes provide tentative, session-level signals that influence subsequent rankings; explicit feedback takes priority. Amazon Nova interprets revision requests into structured attributes, while deterministic rules select products within the shopper’s budget and garment requirements.

Preserve decisions through revisions. Each revision targets the requested pieces while retaining the other selections. Alternatives become the accepted outfit only when the shopper chooses them. At the shopping handoff, product identities, pictured colors, and chosen sizes are matched to merchant variants so the cart reflects the selected look.

Challenges we ran into

Building complete outfits took sequencing. Garment generation required multiple calls, and their order affected how the pieces appeared together—including whether a blouse looked tucked into trousers. Finding a workable sequence took testing, while serial calls introduced latency into an experience built around quick comparisons.

Accessories depended on the photo. Necklace placement was sensitive to pose and neck alignment. A photo that worked for clothing previews could still fail when applying an accessory.

Judging style required another perspective. Fashion isn’t my area of expertise, which made recommendation quality harder to assess. That partly inspired sharing: shoppers can bring someone they trust into the decision and use that feedback to refine the look.

Accomplishments that we're proud of

Delivering the outfit bake-offs. I brought the original idea into a complete experience: compare outfits on your own photo, advance a favorite, and use choices and feedback to shape subsequent suggestions.

Making scoped agentic commerce work. Calliope turns a short brief and follow-up requests into recommendations from one store’s catalog. That bounded scope connects personalized suggestions and targeted revisions to actual products the shopper can purchase.

What we learned

Mobile design forced a clearer decision hierarchy. A desktop layout could show two outfits, individual garments, and prices together. On a phone, that competed with making the photos readable. I redesigned the comparisons around the immediate decision—“Do I like this look?”—while Calliope maintained the constraints and the detailed pieces-and-prices review came afterward.

Smoke-test the intended API experience early. A successful individual call doesn’t establish that a complete workflow will work. Testing garment combinations, generation order, and photo sensitivity upfront reveals constraints that should shape the product before the interface is built around them.

What’s next for Calliope

Bring outfit bake-offs to retailers. Package the comparison experience as an embeddable SDK, starting with a Shopify catalog integration. Retailers could offer personalized outfit previews and preference-driven recommendations within their own storefront.

Start with shared inspiration. Let shoppers explore community-created outfits, preview them on themselves, and adapt individual pieces to their preferences and budget.

Complete the look. Extend accessory support to bags and bracelets, with testing for placement and consistency across different photos.

Built With

• amazon-bedrock
• amazon-nova
• amazon-transcribe
• livostyle-mcp
• model-context-protocol
• python
• react
• typescript
• vite
• youcam-ai-facial-color-tones-analyzer
• youcam-virtual-try-on