GPT Image 3
Conversational workspace for AI image generation and photo editing
About GPT Image 3
GPT Image 3 is a browser workspace for generating and editing images through conversation rather than through a prompt box that forgets what you asked for. You start from a written idea or a reference photo, pick a result, then keep refining it in the same thread, with the original brief, the source images and the chosen output all still attached. That continuity is the product: most image tools treat every generation as a fresh request, which means a small correction to a good result costs you the context that produced it.
The naming deserves a note before anything else. GPT Image 3 is an independent service and says so on its own model page: "OpenAI, Google, and their model names are the property of their respective owners." It is not an OpenAI product. What it actually does is put three engines behind one interface, each card naming its developer plainly: GPT Image 2 from OpenAI, and Nano Banana Pro and Nano Banana 2 from Google, corresponding to Gemini 3 Pro Image and Gemini 3.1 Flash Image. If you came looking for an official OpenAI release, this is not it, and the site does not claim otherwise.
Describe what you want in plain language, or upload a reference photo to start from. The workspace returns options and you choose one. From there you continue the conversation: ask for a different crop, a change of background, another channel format, or a correction to a detail, and the agent works from the result you already picked rather than starting over. Reference images can be attached at any point to steer style or subject. Finished results are exported for the channel you need.
- •Model choice in one place - Three engines behind one interface, each attributed to its developer, so you can compare an OpenAI model and two Google models on the same brief without three accounts.
- •Conversational refinement - Follow-up edits operate on the result you chose, keeping prompt, source photo and output linked instead of discarding context between generations.
- •Reference-based creation - Start from an existing photo rather than text, which matters for product shots where the object has to stay recognisable.
- •Photo editing alongside generation - The same workspace handles edits to images you bring, not only images it invents.
- •Channel variations - Adapt one source image into several output formats for different placements.
- •Credit-based plans - Usage is metered in credits rather than seats, so an occasional user is not paying for a team licence.
Useful for marketers producing social and ad creative who need several variations of one visual, ecommerce sellers turning product photography into campaign images, and designers who want to compare model output before committing to a pipeline. Anyone who has lost a good result to a follow-up prompt will recognise the problem it solves. Two things to know before you sign up: the homepage countdown labelled a 30-minute offer window resets on every page load, and the showcase gallery is disclosed on the page itself as reference examples that are not outputs generated by this product.
Pricing
$19.90 - $199/mo
- Basic$19.90/mo
- Pro$49.90/mo
- Business$99.90/mo
- Max$199/mo
From the vendor pricing page, 2026-09-22
