An image model that follows a long description better than most and can write text in the frame. Good for ad banners, covers and scenes where the result has to match exactly what you described. It draws from scratch and edits your photo.
Create an image with GPT Image 2

| What it does | Image from a description and photo editing |
|---|---|
| Sizes | 1024×1024, 1024×768, 1024×1536, 1920×1080, 2560×1440, 3840×2160 |
| Quality | Low, medium, high |
| Price per image | $0.01 to $0.53 |
| Photo edit price | $0.02 to $0.54 |
| Content filter | Yes |
A headline, a price or a button on a banner comes out readable. For ads this is the main advantage: no need to add the text in an editor afterwards.
You can list ten details at once: background, clothes, light, composition. The model keeps most of them instead of clinging to the first two.
Low costs about a cent and is enough to test an idea. Medium and high give more detail and cleaner text and cost noticeably more.
Six ready sizes from 1024 pixels up to 3840×2160: landscape, portrait and square. The price depends on size and quality.
The same model changes the background, clothes and details on an uploaded photo. You can keep the result at the size of the source.
The newer model holds the description more closely, writes cleaner text and supports large sizes up to 4K. Version 1.5 has only three sizes and no 4K.
Take low to test an idea: about a cent. Make the final banner with text on medium or high, where the letters come out cleaner.
Yes. Editing has a “same as source” option, and the proportions of the frame stay unchanged.
This model has a content filter and may reject real public figures. There is no such limit for your own photos and fictional characters.
No. You pay for each image at the price shown before you run.