# Stanford student turns his photo gallery into personalized virtual wardrobe app

*Published:* 2026-07-14
*Author:* Farzan Hussain

![](https://bestforandroid.com/wp-content/uploads/2026/07/Thijs-Virtual-Wardrobe-App-powered-by-GPT-5.6-Sol_tops.jpg)

A full-stack dev gave an AI model something most of us would never hand over: his entire camera roll. It sorted his whole clothing collection, organized it, and turned it into a personalized virtual wardrobe app with proper categorization and detailing. The final output is the kind of thing that will make you want to try it yourself.

Thijs Simonian, a full-stack developer and Stanford student, gave OpenAI’s new GPT-5.6 Sol model access to his camera roll. Sol pulled out photos containing every piece of clothing he owns, then used GPT-Image to generate new outfit combinations and render them onto his own body. He posted the results five days after [Sol’s public release](https://bestforandroid.com/radar/openai-gpt-5-6-sol-launched-stuck-behind-us-government-approval/), and it’s already circulating as one of the more talked-about builds from the new model’s launch window.

Picture your own camera roll for a second. Old selfies, screenshots you forgot to delete, a blurry photo of your dog from three years ago, and probably a few things you’d rather nobody dig through. That’s what an AI model would be sorting through, looking for shirts and jackets while it’s in there.

> i gave 5.6 sol access to my camera roll and had it extract pictures of every piece of clothing i own from my photos  
>   
> then, told it to find new outfits for me and render them on me with gpt-image!  
>   
> its kinda cool to see your entire wardrobe in a collection like this <https://t.co/pkLTjtn7xL> [pic.twitter.com/SV796uScrB](https://t.co/SV796uScrB)
> 
> — Thijs (@cdngdev) [July 13, 2026](https://x.com/cdngdev/status/2076812846793650485?ref_src=twsrc%5Etfw)



What GPT-5.6 Sol actually pulled off
------------------------------------

Sol didn’t just find pictures with clothes in them. It identified which items were unique across dozens of photos, isolated them, and rebuilt each one as its own clean image. Then GPT-Image took those pieces and generated new outfit combinations, rendered on Simonian’s own likeness, so he could see how they’d look together before actually wearing them.

![Thijs Virtual Wardrobe App powered by GPT-5.6 Sol outfits](https://bestforandroid.com/wp-content/uploads/2026/07/Thijs-Virtual-Wardrobe-App-powered-by-GPT-5.6-Sol_outfits.jpg)

Model releases like this rarely come with just one flashy demo. GPT-5.6 Sol shipped five days ago as OpenAI’s new agentic model, built to run for long stretches on its own using tools and skills without much hand-holding.

That’s exactly why builds like this one are showing up so fast. Developers are testing what a model can do when it’s left alone with a task for a while.

![Thijs Virtual Wardrobe App powered by GPT-5.6 Sol accessories](https://bestforandroid.com/wp-content/uploads/2026/07/Thijs-Virtual-Wardrobe-App-powered-by-GPT-5.6-Sol_accessories.jpg)

Why handing over a full camera roll is the risky part
-----------------------------------------------------

Simonian gave the model access to everything. Not a folder of outfit photos, the entire camera roll. That means the AI touched every screenshot, every private photo, every accidental shot you took and never deleted.

Once that data passes through a company’s servers, you’re trusting its retention policy and training practices, whatever they turn out to be. A wardrobe app is a fun idea. An app that needs your whole photo library to work is a different conversation.

The version of this that actually makes sense
---------------------------------------------

Simonian [shared the underlying method](https://gist.github.com/tandpfun/b73063c8be8fc46644da9925d48b3240) along with access to the SKILL.md file so other developers and Sol users could try it too. It works by scanning a folder of photos for people wearing clothes, isolating each unique garment across those images, creating categories, and reconstructing them into a clean, background-removed cutout that looks like a product shot.

Nothing about that method requires your whole camera roll. Point it at a folder of five or ten photos where you’re wearing different outfits, and you get the same clean wardrobe collection without ever letting a model near your screenshots.

The idea is good. Handing over your whole camera roll to get there isn’t.