A billboard of a woman in a black turtleneck above a Manhattan street at dusk

A billboard of a woman in a black turtleneck high above a Manhattan avenue at dusk, the Empire State Building lit behind it.A billboard of a woman in a black turtleneck high above a Manhattan avenue at dusk, the Empire State Building lit behind it.

Generate

GPT Image 2. The brief, to the letter.

OpenAI’s image model, in your workspace. Write the brief with the words in it, add up to ten references, and get 2K stills and layouts that follow it, in nine shapes.

Generate with GPT Image

Free to start. No credit card.

RESOLUTION
2K
every take
QUALITY
3 levels
Low · Medium · High
ASPECTS
9 ratios
from 16:9 to 9:16
REFERENCES
Up to 10
image references

Built for briefs with words in them.

Headlines, layouts and faces at 2K, in the shape the placement needs.

A freckled woman smoothing cream onto one cheek
LowMediumHigh

Three quality levels

Low for a fast first pass, Medium or High when the take is going out.

A woman with silver makeup among floating lilac cosmetics
16:94:59:16

Nine ratios

From a 16:9 banner to a 9:16 Story, every take at 2K.

A woman in a red dress holding a perfume bottle in the desert at sunset

Four takes a press

Ask for up to four outputs from one prompt and keep the strongest.

Posters with the copy set.

Headlines, taglines and the small print come back as real type, laid out where a designer would put them.

This poster was made with GPT Image 2.

Generate with GPT Image
A beauty launch poster reading Beauty Redefined beside a woman in pink metallic makeup

Nine shapes, one brief.

Pick the placement before you press: a 16:9 banner, a 4:5 post or a 9:16 Story. Every take comes back at 2K.

Generate with GPT Image
A woman in black with long copper hair under red canopies
16:9Banner
A woman in black with long copper hair under red canopies
4:5Post
A woman in black with long copper hair under red canopies
9:16Story

Edit with a sentence.

Attach an image and say what should change. GPT Image 2 can restyle it, recompose it or turn it into something else entirely, like a front page.

This edit was made with GPT Image 2.

Edit with GPT Image
A woman taking a selfie with a brown bear on a forest path
The selfie printed on a newspaper front page under the headline Forest Selfie Goes Viral

Your photo

Layouts, not just pictures.

Ask for the social posts, the lookbook page and the product card, and they come back laid out, frames and all.

These were made with GPT Image 2.

Generate with GPT Image
Social posts, a lookbook page and a product card for a dark fashion collection

How to use GPT Image 2

Three steps in Image Generate.

Generate with GPT Image
A woman in a patterned jacket against a bright graffiti wall

Introducing GPT Image 2

Generate with GPT Image

OpenAI’s image model for briefs with words in them: set type, full layouts and edits from a sentence, at 2K in nine shapes.

More from the workspace

A cream classic car with its hood open in the desert

Image Edit

Change an image you already have

The Blue Mosque and the Bosphorus in Istanbul at sunset

Canvas

Frames and images on one board

A man in sunglasses and a brown shirt against a brick wall

Remove Background

Cut the subject out of an image

A portrait at sunset, sharpened

Upscale

Sharpen an image up to 4K

FAQs

OpenAI’s image model for generating and editing stills. In MarketingX it runs in Image Generate and Image Edit, on your plan.

Every take is 2K. The Quality chip sets Low, Medium or High; Low is the default.

Nine: 16:9, 3:2, 4:3, 5:4, 1:1, 4:5, 3:4, 2:3 and 9:16. 9:16 is the default.

Up to ten reference images, and up to four outputs per press.

5,000 characters. Prompt enhancement is a switch, off until you turn it on.

No. Write what you want to see, and press Generate again for another take.