The ChatGPT Capsule Wardrobe Prompt (Using Only Clothes You Already Own)
Ask ChatGPT to build you a capsule wardrobe and it will almost always hand you a shopping list. A white shirt you do not own. A camel coat you do not own. Straight leg jeans in a wash that is not the wash you have. The output looks beautiful and costs about eleven hundred dollars to act on.
That is not ChatGPT being unhelpful. It is ChatGPT doing exactly what it was asked. "Build me a capsule wardrobe" is a request to design an ideal, and an ideal is easier to design from scratch than from the actual contents of your actual wardrobe, which ChatGPT has never seen.
This page is the other version. The prompt below builds a capsule from what is already hanging in your closet, refuses to name anything you have not listed, and treats a gap as information rather than as an invitation to buy something.
The short version
A capsule wardrobe prompt only works if three things are true. ChatGPT has a real list of your clothes rather than a vibe. The prompt explicitly forbids introducing items you did not name. And the output is measured in outfits you can wear this week rather than in pieces you should acquire. Get those three right and ChatGPT is a genuinely good capsule builder. Get any of them wrong and you get a mood board with a receipt attached.
Why every capsule wardrobe prompt online gives you a shopping list
Search for a capsule wardrobe prompt and you will find dozens. Almost all of them share the same shape. They ask for your body type, your colour season, your lifestyle and your budget, and then they ask ChatGPT to produce a thirty piece capsule.
Budget is the tell. A prompt that asks for a budget has already decided the answer involves spending. And because the prompt gave ChatGPT nothing about your existing clothes, the model has no choice but to invent the capsule from general knowledge about what a good capsule contains. General knowledge about capsules is essentially a list of neutral basics. So that is what comes back.
There is a second reason, and it is less obvious. Capsule wardrobe content is a commercial category. The pages that rank were mostly written by people with affiliate links, and the training data ChatGPT learned from is full of that content. When you ask for a capsule, you are pulling on a thread that is tied at the other end to a shopping list. You have to cut the thread deliberately.
What a capsule wardrobe actually is
A capsule wardrobe is not thirty three items. It is not a colour palette. It is a set of clothes with a high combination rate, meaning most tops go with most bottoms and most shoes go with most outfits, so a small number of pieces produces a large number of wearable looks.
That definition matters because it changes what you are asking ChatGPT to do. You are not asking it to curate taste. You are asking it to solve a combinatorial problem: given this specific set of garments, which subset produces the most wearable outfits for the life I actually live?
That is a task ChatGPT is good at, as long as it can see the set.
Before you paste anything: the twenty minute inventory
The prompt will not work on a vibe. It needs a list. Twenty minutes with your phone is enough for a first pass.
Go through your closet and write one line per item in this shape:
Category | Colour | Fabric or texture | Fit or cut | Formality 1 to 5 | Notes
For example: Trousers | Charcoal | Wool blend | Wide leg, full length | 4 | Needs a heel or they drag
Do not photograph anything yet. Do not try to be complete. Cover the things you would actually consider wearing in the next three months and skip the ski jacket. Aim for forty to eighty lines. Under thirty and the capsule will be thin. Over a hundred and you will start hitting the limits described in Why ChatGPT Forgets Your Wardrobe.
If you would rather work from a structured file than a typed list, the column layout in The ChatGPT Wardrobe Inventory Template is built for exactly this and parses more reliably.
The prompt
Paste your inventory first, then this prompt directly underneath it in the same message. The order matters. Putting the list first and the instruction second keeps the instruction closest to the point where the model starts writing.
You are building a capsule wardrobe from the wardrobe list above.
HARD RULES
1. You may only use items that appear in the list above. If a piece
is not on the list, it does not exist. Do not name it, do not
imply it, do not describe an outfit that requires it.
2. Do not recommend anything to buy anywhere in this response.
If you think a gap exists, note it at the very end under a
heading called OBSERVED GAPS and say nothing more about it.
3. Do not describe my body, colouring or "type". Work only from
the garments.
MY CONTEXT
- Where I actually spend my week: [e.g. 3 days office, 2 days home,
weekends outdoors, one evening thing a fortnight]
- Climate right now: [e.g. Sydney, late winter, 12 to 19 degrees]
- What I need to feel in clothes: [e.g. covered, unfussy, not
corporate]
- Things I will not wear no matter how well they work:
[e.g. anything cropped, heels above 4cm]
WHAT TO PRODUCE
STEP 1. CORE SET
Choose 12 to 18 items from my list with the highest combination
rate. For each one, give a single line saying why it earned its
place, referring to the specific other items it connects to.
Rank them. The item at position 1 should be the one whose removal
would cost me the most outfits.
STEP 2. THE OUTFITS
Build every workable outfit from the core set. Not a sample. All
of them. Group them by occasion using my context above. Write each
outfit as a single line listing the exact items by their list
wording. Number them continuously.
STEP 3. THE BENCH
List the items from my wardrobe that did NOT make the core set,
and sort them into three groups:
- Works, but only in one or two outfits
- Would work if paired differently, and say with what
- Does not combine with anything else I own, and say why
STEP 4. THE WEEK
Using only the core set, lay out seven days that match my actual
week described above. No repeats of a full outfit.
STEP 5. OBSERVED GAPS
Only now, and in under 80 words, name the connections my wardrobe
cannot make. Describe them as functions, not as products.
Example of the right register: "you have nothing that bridges the
formality gap between your knitwear and your tailoring."
Do not name brands, prices or shopping categories.
FORMAT
Plain text. No preamble. Start at STEP 1.What each part of the prompt is doing
The hard rules are first and numbered. Instructions placed at the start of a block and given a hard label are followed more reliably than the same instruction buried in a paragraph. Rule one is the one doing the real work. It converts "build a capsule" from a design task into a selection task.
The ban on shopping is separate from the gap section. If you only say "do not recommend purchases", ChatGPT will still salt the outfit descriptions with implied purchases. By giving gaps their own late, short, tightly worded section, you give the model a legitimate place to put that impulse so it stops leaking into everything else.
The rule against describing your body is not politeness. Once ChatGPT starts reasoning about body type or colour season it drifts towards generic advice, because that is where the bulk of its training data on the subject lives. Keeping it focused on garments keeps it focused on your garments.
Step one asks for a ranking, not a set. A ranked list forces the model to compare items against each other rather than produce a plausible group. The instruction about which item's removal would cost the most outfits is doing the mathematics of a capsule out loud.
Step two says all of them, not a sample. Left alone, ChatGPT will give you five outfits and offer to continue. Asking for the complete set is what turns eighteen items into the forty or fifty looks that make a capsule feel worth it.
Step three is the section people skip and should not. The bench tells you what is dead weight in your wardrobe and, more usefully, what is nearly working. "Would work if paired differently" is where most people find two or three outfits they had never considered. If that section interests you more than the capsule itself, the longer version of it is The ChatGPT Wardrobe Audit.
What good output looks like
You will know it worked if the outfits are boring in a specific way. Real outfits from a real closet are not editorial. They are the grey knit with the wide leg trousers and the flat boots, listed plainly, and you read it and think yes, that is a Tuesday.
You will also know it worked if the core set includes at least one item that surprises you. A capsule built correctly usually promotes something you undervalue, because combination rate and personal fondness are different measurements.
Count the outfits. Twelve to eighteen items should produce somewhere between thirty and sixty workable combinations. If you got fifteen, the prompt was not followed and you should ask for the complete set again.
The three ways this prompt fails
It invents items. You will see a piece appear that is not on your list, usually something so ordinary that the model assumes everyone owns one. A plain white tee. A denim jacket. The fix is one line: "Item 4 in outfit 12 is not on my list. Rebuild every outfit that used it, using only listed items."
It quietly ignores half your list. This is the more common failure and the harder one to spot, because the output looks complete. It happens when the list is long. The fix is to ask directly: "List every item from my original wardrobe list that you did not mention anywhere in your response." If that list is long, your inventory exceeded what the model held onto, which is a context problem rather than a prompt problem.
It drifts back to shopping by message three. You ask a follow up question, and the answer contains a suggestion to pick up a pair of loafers. Memory of the hard rules degrades as the conversation gets longer. The fix is to restate rule one and rule two at the top of any follow up message that matters.
The follow up prompts
Once the capsule exists, these five are worth having. Each one assumes the previous output is still in the conversation.
For a single day rather than a whole capsule, the shorter method in How to Get ChatGPT to Pick Your Outfit is the one to use.
Take outfit 7 and give me three variations of it using only listed items, changing one piece at a time.Which single item on my bench would move to the core set if I changed how I styled it? Show the outfits it would unlock.Rebuild the core set for six months from now, same climate shift, same list.My week has changed: [describe]. Rework STEP 4 only.Give me the three outfits from this capsule that I am least likely to have thought of myself, and say why each one works.
When this prompt stops being enough
The prompt works. It has one structural limit, and it is worth being straight about it rather than selling past it.
Every time you open a new chat, the wardrobe is gone. You paste the list again. You paste the prompt again. You get a capsule that is slightly different from last time, because the model is re-solving the problem from scratch and there is nothing carrying your decisions forward. Nothing knows that you already rejected outfit 12, that you wore the charcoal trousers three times last week, or that you bought a jacket in April.
That is not a flaw in the prompt. It is what a chat window is. ChatGPT holds a conversation, not a database. The mechanism is explained properly in Why ChatGPT Forgets Your Wardrobe, and the four ways of getting around it are compared honestly in Four Ways to Tell ChatGPT What Is In Your Closet.
The fix, when the paste-every-time loop gets tiring, is to give ChatGPT a wardrobe it can query rather than a list it has to hold. That is what Modebase does. Your clothes live in a structured wardrobe, ChatGPT connects to it, and the capsule prompt above becomes a question you can ask in one line because the list is already there. Same prompt logic. No pasting, and it remembers what you wore.
Until you want that, the prompt above is free and it works. Copy it.
Frequently asked questions
Does the ChatGPT capsule wardrobe prompt work on the free plan?
Yes. Nothing in the prompt requires a paid feature. The one difference is that free plans hold less of the conversation in view, so a long inventory is more likely to be partly forgotten. If you are on the free plan, keep your list under about fifty items and check the output using the "list every item you did not mention" follow up.
How many items should be in a ChatGPT built capsule wardrobe?
Twelve to eighteen is the range that produces the best ratio of outfits to items for most people. Below twelve you run out of variety within a week. Above eighteen the combination rate starts falling because you are including pieces that only connect to one or two others.
Why does ChatGPT keep suggesting clothes I do not own?
Because it was trained on capsule wardrobe content that is almost entirely commercial, and because it fills gaps with the most statistically ordinary garment when it is not explicitly forbidden from doing so. The hard rules at the top of the prompt are what stop it. If it still happens, the rules have fallen out of context and need restating.
Can I use this prompt with photos of my clothes instead of a typed list?
Partly. ChatGPT can identify garments in photos reasonably well, but it can only look at a limited number of images at once and it will not retain them across a conversation the way it retains text. Photos work for ten or fifteen items. For a whole wardrobe, a typed or structured list is more reliable. The trade offs are set out in Four Ways to Tell ChatGPT What Is In Your Closet.
Is a capsule wardrobe built from what I own still a capsule wardrobe?
It is the only kind that has ever made sense. A capsule is defined by combination rate, not by having been purchased together. A capsule assembled from clothes you already wear has a much higher success rate than one bought new, because every item has already passed the test of you being willing to put it on.