Does ChatGPT remember my brand between chats?

It can if you set it up. A Project keeps chats, files and instructions together, so every chat in it starts with your brand notes and reference photos, and project instructions override your general custom instructions. It does not check each image against those notes.

Can ChatGPT make a product photo with a transparent background?

Yes. OpenAI’s help center says ChatGPT Images can make the background transparent when you ask, and the API supports transparent PNG and WebP output.

Why did ChatGPT change my label or logo?

Image models draw the whole picture again instead of pasting your photo in, so small text and logo shapes can shift. Use a sharp, straight-on label photo as a reference, say that the label must match exactly, and put the real label in afterward when it has to be exact. See why AI changes your product.

Is ChatGPT or Gemini better for product photos?

On public leaderboards in late September 2026, OpenAI’s GPT Image 2.5 models rank first for both new images and edits, and Google’s Nano Banana models rank lower. Leaderboards use general prompts, so run the same product shot in both and count how many images you would keep.

Can ChatGPT make product photos?

Checked September 24, 20266 min read

Short answer

Yes. ChatGPT Images 2.5, released on September 8, 2026 and available on every plan including Free, can place an uploaded product in a new scene, swap the background, add text and make a transparent PNG. OpenAI says it keeps subjects from reference photos more faithfully than before. It works best for single images and quick edits; OpenAI’s own docs warn that recurring brand elements can drift across many generations, so check every image against the real product.

What ChatGPT Images 2.5 changed

OpenAI released ChatGPT Images 2.5 on September 8, 2026. It says the model is better at preserving the subjects in your reference photos, follows edits more reliably over several turns, and makes images up to 50% faster than Images 2.0. The release also added Sketch (draw a rough layout as a reference), templates for common formats including product photos, comments you place on an image to change one spot, and prompts you can share.

Images 2.5 is on every ChatGPT plan, on web, iOS and Android. You can pick any aspect ratio with a picker or name it in the prompt, and ChatGPT will make the background transparent when you ask.

How to get a usable product photo from ChatGPT

Start from a good photo of the product: straight on, even light, label readable. See how to photograph your product for AI.
Upload one to three references: the front, plus the label or the back if the shot shows them. Fewer, well-chosen references give the model less to mix up.
Write the prompt in a fixed order: the product and which reference it is, the surface or background with a hex color, the light as it falls on the surface, the colors, one sentence on camera and lens, and what must not change. The template below follows that order.
Set the aspect ratio with the picker or name it in the prompt.
Fix one area at a time. Select the part with the editor and describe the change, or place a comment on the image. OpenAI notes that selections are not always precise and edits can spread past the area you picked.
Check the result at full size against the real product: label text, logo shape, color, proportions and parts. Use the QA checklist.
Save the prompt of every image you keep, and start the next shot from it, changing only framing and pose, so the set stays consistent.

A product prompt for ChatGPT (fill in the brackets)

This is a photograph of the exact product in the reference image: [product name], [material and finish].
It sits on [surface or background], [color name] (#[hex]), with a soft contact shadow under it.
[Light, for example: soft window light from camera left falls across the front label and leaves a gentle shadow on the right side.]
Colors stay exactly as in the reference: [color name] #[hex], [color name] #[hex].
Shot on an 85mm lens at f/8, camera at [height], [framing, for example: centered, product fills about two thirds of the frame].
The label text, logo, shape and proportions match the reference exactly, nothing added and nothing removed.
No extra text, no props touching the product, photorealistic.

On GPT Image models, the prompts behind photos our agency kept ran about 300 words, longer than the 140 to 200 words that worked on Google’s Nano Banana models. More on structure in how to write an AI image prompt for a product.

Where ChatGPT falls short for product work

Sets drift. OpenAI’s developer docs warn that the model can struggle to keep recurring characters or brand elements consistent across many generations. A 12-photo set needs more checking than one hero shot.
Products change a little. In a September 2026 test by OpenRouter, a service that sells access to many AI models through one account, four image models, GPT Image 2 among them, were told to keep a coffee bag and its label exactly and change only the background. All four kept the words and the typeface; all four changed the bag’s shape, proportions or lighting to some degree.
Text placement. OpenAI says the model can still struggle with precise text placement, so long headlines and legal copy belong in a design tool. See how to get text right in AI images.
Brand rules are not enforced. A ChatGPT Project can hold your brand notes, reference photos and instructions for every chat inside it, but nothing compares each image with those rules. That check is yours.
Volume. The Free plan has limited and slower image generation. Pro is listed as unlimited, subject to abuse guardrails.
Training. On Free, Go, Plus and Pro, OpenAI may use your content to train its models unless you opt out in settings.

Plans and what they include for images

Plan	Images, in OpenAI’s words	US App Store price a month
Free	Limited and slower image generation	$0
Go	More image creation	$8.00
Plus	More complex and accurate image creation; projects	$19.99
Pro	Unlimited and faster image creation, subject to abuse guardrails	$100 (Pro 5x) or $200 (Pro 20x)

Image lines from OpenAI’s pricing page and prices from the US App Store, September 24, 2026. OpenAI does not publish image counts per plan, and its web pricing page shows prices in local currency.

Using the API instead of the chat

For batches, the same models are in OpenAI’s API as GPT Image 2.5 Flare, the faster default, and GPT Image 2.5 Sunburst, which is slower and meant, in OpenAI’s words, for "polished product imagery". Both cost the same per token as GPT Image 2: $8 per million image-input tokens and $30 per million image-output tokens. Images can be up to 3,840 pixels on the long edge, backgrounds can be transparent in PNG or WebP, and the lowest usage tier allows 5 images a minute.

What one image costs depends on the quality setting. OpenRouter measured GPT Image 2 at $0.006 for a square image at the default low quality and $0.21 at high. Artificial Analysis lists both 2.5 models at about $0.21 an image at their top "max" setting. Compare models in which AI image model is best for product photos.

Can you use ChatGPT product photos commercially?

OpenAI’s Terms of Use say you own the output and that OpenAI assigns you any rights it has in it. They also say output may not be unique, so another user could get something similar. Images from ChatGPT carry C2PA metadata and an invisible SynthID watermark that mark them as made with OpenAI tools, though the metadata can be stripped when a file is edited or converted.

Owning the file does not settle whether it can run in an ad. That depends on what the image shows: an accurate product, no other brand’s marks, no real person’s likeness. See can you use AI images commercially and what Content Credentials are.

Questions people also ask

Does ChatGPT remember my brand between chats?: It can if you set it up. A Project keeps chats, files and instructions together, so every chat in it starts with your brand notes and reference photos, and project instructions override your general custom instructions. It does not check each image against those notes.
Can ChatGPT make a product photo with a transparent background?: Yes. OpenAI’s help center says ChatGPT Images can make the background transparent when you ask, and the API supports transparent PNG and WebP output.
Why did ChatGPT change my label or logo?: Image models draw the whole picture again instead of pasting your photo in, so small text and logo shapes can shift. Use a sharp, straight-on label photo as a reference, say that the label must match exactly, and put the real label in afterward when it has to be exact. See why AI changes your product.
Is ChatGPT or Gemini better for product photos?: On public leaderboards in late September 2026, OpenAI’s GPT Image 2.5 models rank first for both new images and edits, and Google’s Nano Banana models rank lower. Leaderboards use general prompts, so run the same product shot in both and count how many images you would keep.

Where Overs fits

Overs can use GPT Image 2.5 or GPT Image 2 as the model that makes the final photos, as well as Google’s Nano Banana models. Around the model it adds brand rules applied to every prompt, a shot plan that names the reference photos for each shot, and approve, change or reject on each image. The plan is free for 40 photos a month; the AI costs a few cents a photo on your own key from OpenRouter, a service that bills many AI models through one account.

Free for 40 photos a month. The AI that makes the photos is billed separately, on your own key, with no markup from Overs.

Start free How it works

Sources

Checked on September 24, 2026. Prices, specs and rules change; follow the links for the current versions.