# Lewis Muguna

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Lewis Muguna](https://cerebralvalley.ai/u/Lew254)
- **GitHub:** https://github.com/lewis-hue/luxish
- **Demo video:** https://youtu.be/5W_hC0FZsJY?si=iCppqdJmM9yKbdRV
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1728

LUXish is a wardrobe app. You photograph the clothes you own, turning them into something you can actually browse and use. The same pipeline, pointed the other way, produces product photography for the people who make clothes.

Adding a garment takes one photograph and no typing. SAM 3.1 cuts it out of the background; Muse Spark reads it and fills in the category, colour, fabric, and formality; and Muse Image restages it as a catalogue packshot. Photograph a whole outfit and a detector splits it into separate items, so a shirt, trousers and shoes become three wardrobe entries rather than one picture of you.

The stylist answers the question people actually ask their wardrobe. You ask in words or by voice; it answers in words first, then acts. Asked what goes with something, it names pieces of yours and composites them into a figure using cut-outs already on the phone, instantly and at no cost. Asked where to buy something you do not own, it searches the open web per garment and returns shops you can tap. Asked to see a piece you do not own, it draws that piece and puts it on the figure anyway. A photorealistic render happens only when you ask for one.

It is a social product. 152 endpoints across boards, moodboards, follows, reactions, comments, threaded messages, a planner, and a discovery feed that ranks on what a vision model sees in each photograph rather than on hashtags somebody typed once.

For brands, the same pipeline is a studio. A label uploads one product photograph and gets the frames a drop needs: packshot, cut-out, on a model, and in a scene. Each is an edit of their own photograph, so the weave, the print and the true colour survive rather than being replaced by a plausible picture of a similar product. Brand Studio takes a brief, a cast, and an optional reference look, and returns several ordered frames in a single request. That is the part that makes this a business rather than a feature, because a product shoot costs a little money, real money.

Monetisation is built. A free tier of 12 AI runs per month, and a paid tier of 60, metered per calendar month with rollover; every model call is recorded against the user who caused it. The payment provider is the one remaining integration.

## More from Meta Global AI Developer Hackathon

- [gwenlake](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1725)
- [NirmanOs](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1726)
- [The Ground](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1727)
- [SEO Noobs](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1729)
- [ALBA](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1730)
- [VsionX](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1731)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1728. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt