Context

I grew up watching Totally Spies, Powerpuff Girls and Clueless. These worlds showed me that gadgets could be pink, chunky and slightly ridiculous. Fashion, tech and web design were the three things that pulled me into design in the first place. This project was there to tie those interests together and honor my old tween interests.

Problem

I own plenty of clothes but still feel like I have nothing to wear. This is due to two things I couldn’t do with my own wardrobe: see all of it at once, and find new combinations in it without buying more.

Constraints

Solo build, no team, no budget. The tool had to be useful to exactly one user (me), which meant I could skip generalising and optimise for real daily use. I gave myself a $10 budget for fal.ai to set up a working AI generation pipeline.

Process

I drafted the concept in Claude and turned it into a PRD prompt for Claude Code. I built the scaffolding logic first: rotating through tops, bottoms and shoes, with items tagged by brand, season and description. Shuffle came next, so I could remix a look in seconds. Saving worn outfits turned the tool into a record of what I actually reach for.

Finally, I built a preview function by connecting fal.ai to Claude Code. I generated a base body from a portrait of myself and wrote a prompt that dresses that base using Nano Banana. This way the “dress me” button returns a look on a body, not a flat lay. Seedance handled the low-poly loading animation.

Only once it worked did I touch the UI. I skinned it like an old Winamp player, with Web 2.0 gloss and Frutiger Aero gradients.

Outcome

A working wardrobe I can browse, shuffle and wear-log, and a growing dataset on my own style. Shopping stops being the default answer to “what do I wear.”