ChatGPT Images for Product Mockups: Prompts and Export

Product mockups are one of the most practical uses of image generation in ChatGPT. A brand owner can see a label on a jar before printing it; a designer can show three packaging directions in an afternoon; a shop can produce lifestyle imagery without a photo shoot. The GPT Image model is well suited to this because it follows detailed briefs, renders label text legibly and can output a clean PNG with a transparent background. This guide covers prompting for believable products, keeping the same product consistent across a set and getting files out at a usable size.

Describe the product like a spec sheet

Products fail in AI images when they are vague. A "skincare bottle" could be anything; a "frosted glass 50 ml dropper bottle with a matte black cap and a cream paper label" is a thing the model can draw consistently. Cover these points in order:

1. Form and size cues. Shape, proportions, closure type, anything that signals scale (a hand, a coin, a tabletop).
2. Materials and finish. Frosted glass, brushed aluminium, kraft card, soft-touch matte, glossy lacquer. Finish drives how light behaves, which is what sells realism.
3. Label and text. The exact words in quotation marks, the type style, the placement. Keep it to a brand name and one or two lines. See text and typography in ChatGPT images for the detail.
4. Colour. Name colours plainly. If you have a brand hex value, state it; the model treats it as a strong hint rather than an exact match, so verify afterwards.
5. Scene and surface. Where the product sits and what is near it. Fewer props look more professional.
6. Light and camera. Softbox from the left, slight top light, shallow depth of field, eye-level three-quarter view. Studio language works.
7. Orientation and background. Square for shop listings, portrait for social, landscape for banners; plain, scene or transparent background.

Avoid real brand names and logos other than your own. If you are mocking up a client’s product, use their actual name and have the right to do so.

Three mockup types and how to prompt them

• Packshot on plain background. "On a plain white background, soft even studio light, subtle contact shadow." The standard e-commerce image. Ask for a transparent background instead if you will place it in a layout yourself.
• Lifestyle scene. The product in context: on a bathroom shelf, in a hand, on a café table. Keep the product large in frame and in sharp focus, with the scene softly blurred behind it, or the label will be too small to read.
• Flat lay or packaging spread. Box, inner tray and product arranged from above, overhead camera, even light. Describe the arrangement explicitly or the model will decide for you.

Keeping the product identical across images

A product that changes shape between the packshot and the lifestyle image undermines the whole set. The reliable method is a master image plus a fixed description.

• Generate the product once on a plain background and iterate until it is right. Save that image as your master.
• Attach the master to every subsequent prompt with: "The product in the attached image is the exact product to show; keep its shape, materials, colours and label identical."
• Paste the same product description, word for word, into every prompt.
• If you have photographs of a real product, attach one of those instead; a photo is the most reliable reference of all.
• Use follow-up edits for variations ("same scene, swap the surface to pale marble") rather than fresh generations, so the product carries over.

More on reference-driven consistency in ChatGPT image edits and references.

Settings at a glance

Exporting and upscaling

ChatGPT delivers the image as a downloadable file; ask for PNG when you need transparency. The output resolution is set by the service and varies by plan, and it is generally enough for screens and social media but not for large print. For anything that will be printed or shown large, run the file through a dedicated upscaler afterwards. Most upscalers handle product shots well because the surfaces are smooth and the edges clean, but they can soften label text; if that happens, upscale the image and then re-set the small text in a design tool over the top.

A sensible export checklist: download the full-size file rather than a screenshot; keep the original PNG untouched and work on copies; name files with the product, variant and version (serum-50ml_white-bg_v3.png); and store the prompt and the master reference alongside the file so you can regenerate a variant later. For the current output sizes and download options, check OpenAI’s help pages.

Try it

Create a square product photograph of a frosted glass 50 ml dropper bottle with a matte black cap and a cream paper label. The label reads "Linden & Co" in a thin serif, with "Night Serum" below in small capitals; no other text. The bottle stands on a pale oak surface against a plain warm grey background, soft studio light from the left with a gentle fill, subtle contact shadow, shallow depth of field, eye-level three-quarter view. Clean, minimal, premium skincare photography.

Replace the product sentence and the quoted label text with your own, then keep the lighting and camera sentences as a template. For a cut-out, change the background sentence to "transparent background, output as PNG". For a lifestyle version, attach this result as the reference and ask for the same bottle on a bathroom shelf with soft morning light.

Common problems

• The label text is wrong. Shorten it, quote it, and fix any remaining error with a follow-up edit.
• The product changes between images. You did not attach the master or you paraphrased the description. Attach and paste verbatim.
• It looks like a render, not a photo. Add imperfections: a slight reflection, a soft shadow, a hint of dust on the surface, real-world materials.
• Too many props. Ask for a minimal scene with one or two supporting objects at most.