AI News

GPT-6 Astra outfit workflow: upload a few photos to generate a full year of OOTD

36Kr · AIUpdated Nov 16 19:55

AI summary

AI blogger Rowan Cheung built a personal outfit workflow with GPT-6 Astra. All you need to do is upload a few photos, provide your height, weight, and clothing preferences, and it can generate a full year of OOTD, producing 30 outfit plans per minute. In testing, GPT-5.6 sol was enough to complete it; the author says 5.6 finished in one hour work that 6 Astra hadn't finished after waiting a whole day. The workflow can also read real-time weather to pick the day's outfit, tag how many sets each item can be paired with, and sort clothes that can't be combined into a "to be cleared" category.

Body · AI translation

What does it take to be an outfit blogger?

A mirror, a wardrobe, and lots of nice clothes.

Now, GPT-6 tells you that none of these are needed anymore...

Recently, AI blogger Rowan Cheung built a personal outfit workflow using GPT-6 Astra.

Just upload a few photos, report your height, weight, and clothing preferences, and it can generate a full year of OOTDs.

At first glance, you'd think it was an autumn new-arrivals catalog from some brand's official website.

The man in the images is all Rowan.

With the corners of his mouth turned up, Rowan expressed his excitement: My colleagues can't complain anymore that I wear the same black T-shirt every day!

That's what he says, but think about it carefully:

What if we took this method to Xiaohongshu to be an outfit blogger?

No need to buy clothes, no need to buy a mirror—just upload photos of a few individual items. In one minute, you can directly generate 30 outfit plans.

Be even bolder—take it straight to e-commerce?

No need to hire a photographer. No need to book a model. No need to rent a venue. The clothes—you don't even need to own them...

In the comments section, Rowan Cheung has thoughtfully attached the complete development prompt and operating procedure.

Step two: copy the long string of prompts Rowan posted at the URL, and also write down the city you live in and your clothing style.

As for performance, testing shows it can be done with GPT-5.6 sol; you don't need to summon a deity once every five hours.

In fact, in the end it was 5.6 that finished in one hour the work that 6 Astra waited a whole day and still didn't finish... This may be the tortoise and the hare race of the AI era.

By this step, it's actually already done.

It's just that the clothing in the page is still generated by GPT itself:

This outfit style, how should I put it... it really makes you wonder whether it was generated based on the region.

It's simply commuter wear tailor-made for us in Haidian.

In any case, you still need to add one more step: send your own clothes to GPT.

Here, to make it easier for AI to extract, I directly used product images from Taobao purchase pages, sending not only several items I'd already bought, but also several I planned to buy for Double 11.

First is the homepage; it reads the real-time weather of the day, then picks an OOTD from the preset outfits.

The entry-function column lists outfits for different seasons, as well as the individual items you need and full-set references, and so on.

For example, recently, in autumn, the outfit it gave me was:

For spring, although the temperature is similar, the colors and visuals should be livelier:

Jumping over to a certain red app to start an account with "What I Wear Every Day in Autumn in Beijing" ing......

Moreover, the items needed for each outfit here are also listed, and each is marked with the number of sets it can be paired with.

—If you have clothes you want to buy but don't know whether they'll suit you once you get them, you can absolutely import them first and see whether they can be paired with your existing clothes.

If there are individual items of clothing that don't make it into any outfit combination, they will be sorted into the "to be cleared" column.

In addition, the website also generates a complete personal outfit profile.

My height, body type, daily usage scenarios, and preferred clothing style will all participate in the later matching.

However.

Although it looks like it already has a lot of features, some netizens still raised their own concerns...

For example: how can it know which piece of clothing should be washed and which is clean?

Someone else gave it a try, and the result—

This isn't much of a problem, because in fact Jobs also only wore the same turtleneck sweater.

However, friends who care about workflow developer Rowan will ask:

Rowan, you wear the same black T-shirt every day, so you never had to agonize over what to wear in the first place. Now that you've used AI to design 30 outfits, can it really reduce your decision burden?

GPT-6's wondrous and marvelous uses

Speaking of which.

I thought there were already enough ways to mess around with AI, but I didn't expect that after GPT-6 arrived, everyone just let themselves go completely, focusing on "whatever you dream of, do it."

Only mortals can fail to imagine it; there's nothing your Astra god can't do.

Someone found Apple Health hard to use, so they just used GPT-6 Astra to spend 13 minutes making their own Apple Health.

Someone wanted to organize their book collection but was too lazy to enter them one by one, so they just took a photo of the bookshelf and had GPT-6 Astra generate a digital bookshelf.

Not only can you search for books, but you can also jump to Goodreads with one click.

Someone used GPT-6 Astra to recreate their apartment and turn daily life into a game:

In the game, she clicks on the notebook and can see to-do items; clicking on books jumps to Goodreads reading records...

She even put her own dog into the game.

No kidding, how is this any different from The Sims?

Someone else used GPT-6 Astra to process a 3D model and 1:1 print their own brain.

Developer Sourish Sharma

Some even connected GPT-6 Astra to a Toyota Corolla and let it drive itself to In-N-Out to buy a burger...

Someone else used GPT-6 Astra to generate a world in Unreal Engine and even put a group of AI residents inside.

Reference links:

This article comes from the WeChat public account "Quantum Bit" (ID: QbitAI), author: Cheng Qian, published by 36Kr with authorization.

Source

36Kr · AI