Outfit Optimizer
With limited space I want a wardrobe that gives me the maximum variety with the least number of items. Ideally everything will compliment one another so I looked into how AI could help with this.
With Claude I built a tool that allows you to input what you own, this did take longer than a couple of hours but was still less than a day. It works out how many outfit combinations you can make from those items. The maths multiplies shirts and knitwear by bottoms by shoes, then layers in jackets (each one doubling your options) and accessories.
The wardrobe is split across six categories: Shirts & T-Shirts, Knitwear, Bottoms, Jackets, Shoes, and Accessories. Items can be added manually or through three preset capsule wardrobes to get started quickly. The outfit counter updates in real time as you build things out.
The two main AI features both run through a Make.com webhook to keep the Gemini API key off the frontend entirely.
I quickly got bored of adding the items manually so introduced a Photo scan. Upload a photo of your wardrobe, a flat lay, or anything with clothes in it. Gemini Vision identifies each item, categorises it, and surfaces them as selectable chips. You pick which ones to add and they drop straight into the right categories. I think it works surprisingly well.
Once you've added the items you have you can then ask for Suggestions on what to buy next. Gemini analyses what you've got, works out where the gaps are, and returns four specific items ranked by how many new combinations each one would unlock. Your also given three colour options for each, chosen to work with what's already there, along with a link to a possible retailer. Ideally I would have the AI give direct links to the suggested products but it was hallucinating too much when generating the links, so a compromise was to link to the retailer with a pre-loaded search url which is more consistent with results.
The Make scenario ended up with a router handling two separate routes one for text-only suggestions using the Gemini Simple Text Prompt module, and one for image scanning using an HTTP module hitting the Gemini Vision REST endpoint directly with the image passed as base64. Getting the two routes to play nicely took a fair bit of back and forth, mostly around how Make parses incoming JSON (flat, not nested under body) and how it handles the Gemini response object.
Its by no means perfect the UI needs further improvement and refinement but as a proof of concept it works well. Initial ideas for future development would be a saved wardrobe and feedback on items you see in store and how well they could work with your existing wardrobe and getting some of my more stylish friends to review the suggestions.
