How Arcade AI works: from one sentence to a finished design
Four steps, no account and about eight seconds of waiting. Step three is a preview simulation, and this page is explicit about exactly where the real generation stops.
The whole flow, before the detail
You write a sentence describing a physical object. Arcade AI reads it and renders a set of design images. If you want to see how one of them would look as a made piece, you send it through the maker preview, which assembles a visualisation in your browser. Then you download whichever image you want to keep. That is the entire tool.
Nothing in that sequence asks for an email address, a password or a payment method, and nothing is stored against your name afterwards. The only gate is a lightweight challenge in front of the generator that keeps automated traffic out. If you close the tab halfway through, nothing is left behind except the images you chose to save.
The four steps
Each step below is a real thing the page does, described in the order you meet it.

Describe the product you want
Type a plain sentence naming the object and two or three details about it — colour, pattern, texture, scale. There is no prompt syntax and no settings panel to learn; the description is the only input the generator takes, apart from an output shape.
Pick square, portrait or landscape from the dropdown next to the button. Portrait suits rugs and wall panels, landscape suits table runners and bed sets, and square is the safe default for everything else.

Arcade AI generates the design
Press generate and the description goes to the design model. In roughly eight seconds a set of finished images appears in place, on the same page. Rewrite a word and run it again as often as you like — there is no counter, no credit balance and no cooldown that pushes you towards a plan.
Iterating by rewriting works better than iterating by settings. Change one thing at a time — the palette, then the scale of the motif — and you learn quickly which word was doing the work.

Preview it as a finished piece(simulated)
Choose a design and send it through the maker preview. It walks through matching, materials, making and finishing, then shows the design as a piece in a room. This step is a simulation shown for inspiration only: nothing is manufactured, produced, ordered or shipped, and no maker is contacted.
The preview runs entirely in your browser and calls no service. Its value is judgement, not logistics — a repeat that looks busy on a flat image often settles down once you see it at room scale.

Download and keep the design
Save the full-resolution image straight from the result card. What you do next is up to you: keep it as a reference, drop it into a mood-board, or hand it to a weaver, potter, printer or jeweller you hire yourself as a brief for the real object.
Download the ones you are unsure about too. Ideas that look wrong on their own often turn out to be the right answer next to the rest of a room.
Where the real generation stops
Steps one, two and four are real. You write a description, a design model renders it, and you download the file that comes back. Step three is not. The maker preview is a visualisation built to help you judge a design in context, and it is labelled as a simulation everywhere it appears on this site, including inside the preview itself.
That distinction matters enough to spell out twice. Arcade AI has no factory, no supplier network and no checkout. Sending a design through the preview does not place an order, does not notify a maker, does not reserve materials and does not start a delivery. If you want the physical object, you take the downloaded design to someone who makes that kind of thing.
Step three is a simulation. Arcade AI visualises how your design would look as a finished piece so you can judge it in context — it does not manufacture, produce, sell or ship anything, and no order is ever placed.
Run the four steps yourself
The generator below is the real thing: type a sentence, wait about eight seconds, download what comes back. No sign-up, no credits, no card.
Four ways to get a better result
Small changes to the wording move the output more than anything else on the page.
Name the object first
Open with the noun — rug, pillow cover, pendant — so the generator knows what shape it is drawing before it reads a single adjective.
Give two or three concrete details
Palette, pattern and texture are the three that change the most. More than four details tend to fight each other rather than add up.
Say the scale of the motif
Words like fine, oversized, repeating or single-panel decide whether a pattern reads across a whole piece or disappears into it.
Change one word, not five
Run the same prompt with a single swapped word to see what that word actually does. It is faster than starting over each time.
Now pick something to design
Every category page opens with a preset prompt in the box and six more you can borrow, so the first run takes no thinking at all.
Create real things with AI — free, right now
One sentence in, a finished product design out. No account, no credits, no waiting, nothing to install.
Free forever · No sign-up · Works in your browser


