20 ChatGPT Image Prompts for Product Photos, One Word Each

November 6, 2026 · 9 min read · by Jan Oršula

The quickest way to get product images out of ChatGPT is one detailed prompt followed by a single word per style. Describe the product once (or upload a photo of it), then send short lines in the same chat: "x-ray this camera", "flat lay of this camera", "movie poster of this camera". The chat remembers the product, so the one word only has to name the look. Below are all 20 words we tested, with the exact picture each one returned, nothing edited or re-rolled.

We ran the whole set on one subject, a vintage film camera, so you can compare the styles side by side. A few of the twenty didn't do what we expected, and those notes are the useful part: they tell you what to add when you run the same prompts on your own product.

The 20 prompts (.md, 2 KB)

The base prompt, all 20 one-word prompts and the fixes from this guide in one file, ready to paste into ChatGPT.

Download the prompts

How the one-word method works

Most prompt lists hand you a paragraph per image: subject, lens, lighting, background, mood. That works, but you rewrite the subject twenty times and it comes out slightly different each time. Splitting the job in two fixes both problems.

The first message sets the subject. Describe the product the way you want it to look in every picture: what it is, colours, angle, background, light. Or skip the writing and upload a clear photo of your real product.
Every message after that changes one thing. "x-ray this camera" works because "this camera" points back at the image already in the chat, and "x-ray" names a visual genre ChatGPT has seen countless examples of. You aren't describing an x-ray. You're asking for one.

Stay in the same chat for the whole set. A new chat forgets the subject and you're back to describing it.

Here's the base prompt we used, and what it returned:

A vintage 1970s 35mm film camera, silver top plate and black leather body, lens attached, three-quarter view, on a plain warm off-white background, soft studio light, square image.

ChatGPT's base image: a silver and black 1970s film SLR camera with a lens, three-quarter view, on a warm off-white background. The top plate reads Asahi Pentax Spotmatic. — The base image. Note the Asahi Pentax logo: we asked for a generic vintage camera, and ChatGPT drew a real model.

See inside: 5 prompts for how-it-works posts

These five open the product up. They suit any post that explains how something works, what's inside it, or why it costs what it does.

The same camera rendered as clear glass, showing the film, gears, prism and lens elements inside. — `x-ray this camera`
The body turns to clear glass and every part shows through. It kept the exact shape and lettering of the base image, so it reads as the same camera.

The camera sliced open, with the prism, mirror, film path and lens glass visible in the cut. — `cross-section this camera`
Sliced open with the inside on show: prism, mirror, film path. Best when the inside is the point of the post.

A collage of three transparent versions of the camera, at different levels of see-through. — `see-through this camera`
A miss: it returned three versions in one collage. Add "one single image" if you want one picture.

Every part of the camera floating apart in order: lens elements, top plate, dials, film door, screws. — `exploded view of this camera`
Every part floats apart, in order. The strongest scroll-stopper of the set, though it gets busy at phone size, so crop in.

A section cut out of the camera's side, edged in red, showing the gears, prism and film inside. — `cutaway of this camera`
A window cut into one side, edged in red. Unlike the cross-section, the outside stays recognisable.

Draw it: 5 prompts for explainers and diagrams

Line drawings and technical sheets. They work as backgrounds for text, as explainer slides, and anywhere a photo would feel like an ad.

The camera drawn as a fine grey 3D wireframe mesh on an off-white background. — `wireframe this camera`
A 3D model made of fine lines. Lovely full screen, muddy as a small thumbnail.

The camera shot from directly above, showing the dials, lens barrel and top plate. — `top-down this camera`
Shot straight down from above. Plain on purpose, and the easiest of the twenty to put a headline next to.

A navy-and-cream blueprint sheet with front, top, side, rear and bottom views of the camera, lens diagrams and a parts list. — `blueprint this camera`
Not one drawing but a whole sheet: six views, lens diagrams and a parts list. The small text is generated, so don't let anyone zoom in on it as fact.

An isometric diagram sheet with the camera, an exploded isometric view of its parts and small detail panels. — `isometric this camera`
A partial miss: we expected one isometric view and got an exploded diagram sheet, in a wider shape. Add "single object, square" to keep it simple.

A black-ink patent-style sheet with nine numbered figures of the camera and a header reading U.S. Patent, Oct. 12, 1976. — `patent drawing of this camera`
Black ink figures with numbered parts, plus a "U.S. Patent" header, date and number that ChatGPT made up. Great look; never post it as a real patent.

The commercial set. If you sell a physical product, start here.

A clean studio photo of a silver and black film camera on a white background. — `product shot of this camera`
A clean studio photo on white. Look closely and the camera has changed: different lens, no Pentax logo. More on that drift below.

A fingertip next to a tiny version of the camera standing on a wooden table. — `miniature this camera`
A tiny collectible version next to a fingertip. The scale trick does the work, so it suits playful launches.

The camera from above on dark wood, surrounded by a spare lens, film canister, lens cap, strap and cleaning cloth. — `flat lay of this camera`
Overhead on dark wood with props it picked itself: film, lens cap, strap, cleaning cloth. The props are what make a flat lay, and it chose sensible ones.

The camera as a matte grey clay render with no colour or text. — `clay this camera`
A matte grey clay render. All colour and branding gone, which makes it handy for concept or "coming soon" posts.

The camera as a die-cut sticker with a thick white border on a dark background. — `sticker of this camera`
Die-cut with a white border. Drop it onto any background in a design app, or use it as a reaction graphic.

Restyle it: 5 prompts that change the genre

These take the product out of product photography entirely. Use them sparingly: one in a carousel stops the scroll, five in a row look like a filter pack.

A clean infographic of the camera with labelled parts, lens details, top, back and side views, and dimensions. — `infographic of this camera`
Labelled parts, three extra views and dimensions. The labels look right, but every spec was generated. Check each word and number before it goes out.

Side-by-side panels labelled Before and After: a dusty, worn camera on the left and a clean, restored one on the right. — `before/after of this camera`
It invented a dusty, worn "before" for a restoration story. Good for cleaning or repair services, as long as you say the before is illustrated.

The camera in heat-camera colours, glowing orange and yellow on dark blue, with a temperature scale from 10 to 60 degrees. — `thermal this camera`
Heat-camera colours with a temperature scale it made up. Pure mood, and very hard to scroll past.

A comic-book page with the camera in five panels, captioned Shutter speed dial, Aperture ring, Back view, Top view and Side view. — `comic of this camera`
A whole comic page, with panels for each part. One prompt, a ready-made carousel cover.

A portrait movie-poster image of the camera on a rock at sunset, with film strips curling behind it. — `movie poster of this camera`
It switched to portrait on its own: sunset, mountains, curling film. There's no title on it, so add your own text on top.

What went wrong, and how to fix it

Twenty prompts, one chat, and five kinds of problem. Each one has a one-line fix you can add to the prompt.

The product drifts after about ten images

Compare the base image with the product shot. Same idea, different camera: the lens lettering changed, the logo vanished, the dials moved. ChatGPT works from its own latest picture, not the first one, so small changes add up. If your real product has to stay exact, open a new chat every eight to ten images, upload the image you started with, and keep going with the same one-liners. For a real product, upload a photo of the real thing instead of describing it.

It drew a real brand we never asked for

"A vintage 1970s 35mm film camera" came back as an Asahi Pentax Spotmatic, logo and all. That's harmless in a guide like this one, but a real company's logo on your promo post is a problem: OpenAI's Terms of Use hand you the rights to what you generate, not to someone else's trademark. Add "no logos, no brand names" to the base prompt, and look at every image for lettering before you post it.

It made up text that looks like fact

The patent drawing has a patent number. The infographic has specs. The blueprint has a parts list. All generated, all plausible, and some of it wrong. When an image carries text, either check every word or add "no text" and put the words on yourself.

The image shape changed mid-chat

The first nine images were square. Then the isometric sheet came back wider, most of the rest at 3:2, and the movie poster in portrait. A one-word prompt gives ChatGPT room to choose. Add the shape you need to the line: "flat lay of this camera, 4:5 portrait".

One prompt, three pictures

"see-through" returned a collage of three versions. Add "one single image" when that happens.

How to run this on your own product

Swap the camera for whatever you sell or post about. Upload a well-lit photo of the product on a plain background and write: "This is my product. Keep it exactly as it is." Then send the one-liners, changing "this camera" to "this mug", "this bottle", "this app on a phone screen".

Not every word suits every product. A rough guide:

Physical products with an interesting inside (gadgets, bags, skincare packaging): x-ray, cross-section, exploded view, cutaway.
Anything you sell online: product shot, flat lay, top-down, sticker.
Services and software, where there's no object: put the idea in the first message ("a laptop showing a content calendar"), then try blueprint, infographic, comic and before/after.
Launches and teasers: clay for "coming soon", miniature and movie poster for the reveal.

When one word isn't enough, add one more detail and stop there: "flat lay of this camera, on white marble" or "comic of this camera, black and white". Two or three words in a line still beats rewriting the whole description.

From picture to post

A picture in a chat isn't a post yet. Three steps get it there.

Fit the shape. Instagram's feed shows portrait posts up to 3:4 (1080 × 1440) without cropping, and most of these came back square or 3:2. Ask ChatGPT for the shape up front, or crop afterwards with the free social media image resizer. The image size guide has the ratio for every network.

Write the caption and the alt text. The words do the selling: a comic-panel camera earns its place when the caption explains how a film camera works, and says little under "new drop". While you're there, describe the image for screen readers; the alt text generator drafts one from the picture.

Schedule it from the same chat. ChatGPT can also put the post in your queue. Kadenzo is a scheduler with a plugin in ChatGPT's Plugin directory: once it's connected, you can say "schedule this as an Instagram post for Thursday at 9 am" in the chat that made the image, check the draft, and confirm. It posts to 11 networks and comes with the Professional plan at $29 a month. The setup takes a few minutes and is in our guide to scheduling social media posts from ChatGPT. One detail to know: images you attach in the chat don't reach plugins, so the guide shows how to give it a public link to the file.

On labels: Meta asks you to disclose photorealistic AI video and audio, and may add an "AI info" label to images it detects as AI-made. For a flat lay or a product shot that people could take for a real photo, saying so in the caption costs nothing and saves the comments section.

Frequently asked questions

Do these prompts work in the free version of ChatGPT?

Yes. Image generation is available on the free plan with a lower daily limit, and the prompts are the same on every plan. On the free plan you may hit the limit before you finish all twenty, so pick the five words that suit your product first.

Why use one word instead of a detailed prompt?

Because the detail is already in the chat. Your first message (or photo) describes the product once; after that, a word like "blueprint" or "flat lay" names a style ChatGPT already knows. You rewrite less, and the product stays closer to the same across the set.

How do I keep my product looking the same in every image?

Start from a photo of the real product rather than a description, say "keep it exactly as it is", and stay in one chat. When the details start to slip, usually after eight to ten images, open a new chat, upload the original photo again and carry on.

Can I use ChatGPT images in ads and posts for my business?

OpenAI's terms assign you its rights in what you generate, so commercial use is allowed as long as you follow its usage policies. Two cautions: AI images may not get copyright protection, and if a picture shows someone else's logo or a real product you don't sell, leave it out.

Which of the 20 styles work best on Instagram?

For a carousel cover, the ones with something to discover: exploded view, x-ray, comic and thermal. For a shop feed, product shot, flat lay and top-down. In our set the line drawings (wireframe, blueprint) looked best full screen and lost detail as small grid thumbnails.