GPT Image 2

Create product images, posters, portraits and illustrations from a written brief or a reference image. Set the subject, style and composition, then refine the details that matter.

Research

Webpage

Slides

Image

Video

Make the brief work harder

Give the model useful direction and make each revision purposeful.

Describe the details that define the image

Generate a new image from text, or add a reference when an existing object should guide the result. Describe material, lighting and composition to turn a visual idea into a concrete image brief.

Illustration of a glass orchid with lighting and framing cues for a detailed image brief

Give an existing image a new direction

Upload a reference image and say what should stay and what should change. Ask for a different setting or lighting while keeping attention on the product, then inspect shape, proportions and fine details in the result.

Illustration of the same product with a new backdrop and light while its shape stays consistent

Refine the part that matters next

A stronger image does not always need a new idea. Adjust the spacing around a headline, simplify the background or change the light. Keep the useful parts of your brief and make the next request specific.

Illustration of a focused adjustment to one detail in a poster

From a brief to your next image

Choose your starting point
Describe a new image or upload a reference. Include the subject, intended use and any details that need to stay consistent.
Open the image task
Choose Try in YouMind to continue with this model, your brief and any reference images. Sign in if prompted; the task workflow determines the generation options available to you.
Review and refine
Inspect the image at a useful size. Check words, product details and faces, then request a focused change before you use the result.

Skills for your next visual project

Explore related workflows for illustrations, picture books and presentation visuals.

Explore related Skills

Browse Skills

Clean lifestyle for reseach/mkt

Slide

Naive Doodle Avatar

Image

Turn Ideas into a Picture Book

Image · Webpage

Bicolor Overprint Effect Presentation

Slide

Compare image models

Both create and edit images from text and references. The table compares published model capabilities; available settings depend on the task.

Model comparison
Compare	GPT Image 2	Nano Banana Pro
Input	Text prompts and reference images, starting from a brief or existing visual.	Text prompts and reference images, starting from a brief or existing visual.
Output	Generated or edited images for layouts, documents and other visual projects.	Images, with provider support for text and illustrations in one response.
Image sizing	Custom pixel dimensions within model limits, for a specific layout.	Aspect ratio selection with 1K, 2K and 4K size options.
Editing an existing image	Reference images with instructions for what to change and keep.	Reference images with instructions for what to change and keep.

Scroll horizontally to compare all models

What our users say

I used GPT Image 2 to explore product photography for a ceramic cup. Describing the glaze, linen surface and morning light gave me a direction I could discuss with the photographer before the shoot.
Maya ReynoldsBrand designer
A jazz night poster needs room to breathe. I started with the headline and two colors, then asked for more space around the lettering. That small revision was more useful than starting over with another style.
LeonEvent organizer
For image editing, I bring a reference and explain what should stay. With GPT Image 2, I can explore a different backdrop for a portrait while keeping the wardrobe and mood at the center of the brief.
Nina KimPortrait photographer
My article was about starting again after a setback. A tiny garden growing through a cracked teacup said more than a generic picture of someone at a desk. I liked being able to develop that metaphor as an editorial illustration.
Owen ParkerNewsletter writer
I'm exploring a courier character for a small game. I use GPT Image 2 for character design studies, changing the coat, bag and silhouette one at a time before deciding what belongs in the final reference sheet.
TessaIndependent game designer
One subject, a quiet background and room for my title. That's the starting point I want for a YouTube thumbnail.
Jules MoreauVideo creator
A kitchen at night, a sleepy fox and a lamp shaped like a moon. For my picture book, I use GPT Image 2 to explore how the setting can carry the feeling before I add any dialogue.
Ada F.Picture book author
Packaging concepts are easier to discuss when I can see the box on a shelf. I describe the paper texture, the color blocks and the viewing angle, then use the image to decide which direction deserves a physical mockup.
Callum FraserPackaging designer
My recipe handout needed a close view of a dumpling fold. I wrote the image prompt around the hand position and the edge of the wrapper. I check those details carefully before using an illustration to teach a real technique.
Esme LaurentCooking instructor
Same reading nook, different mood. I explore warm wood, muted green and softer evening light with a reference image before choosing materials for the room.
Felix BraunInterior stylist
For a slide about fragile systems, I wanted a bridge made from folded paper. GPT Image 2 helps me develop a visual metaphor around the point I am making, instead of settling for an unrelated stock photo.
Imogen ClarkePresentation designer
I plan my social post cover around where the text will go. Asking for the main object off to one side gives me a composition to work with; I add the final wording and check the crop in my publishing layout.
Rhea ShahCommunity editor

I used GPT Image 2 to explore product photography for a ceramic cup. Describing the glaze, linen surface and morning light gave me a direction I could discuss with the photographer before the shoot.
Maya ReynoldsBrand designer
A jazz night poster needs room to breathe. I started with the headline and two colors, then asked for more space around the lettering. That small revision was more useful than starting over with another style.
LeonEvent organizer
For image editing, I bring a reference and explain what should stay. With GPT Image 2, I can explore a different backdrop for a portrait while keeping the wardrobe and mood at the center of the brief.
Nina KimPortrait photographer
My article was about starting again after a setback. A tiny garden growing through a cracked teacup said more than a generic picture of someone at a desk. I liked being able to develop that metaphor as an editorial illustration.
Owen ParkerNewsletter writer
I'm exploring a courier character for a small game. I use GPT Image 2 for character design studies, changing the coat, bag and silhouette one at a time before deciding what belongs in the final reference sheet.
TessaIndependent game designer
One subject, a quiet background and room for my title. That's the starting point I want for a YouTube thumbnail.
Jules MoreauVideo creator
A kitchen at night, a sleepy fox and a lamp shaped like a moon. For my picture book, I use GPT Image 2 to explore how the setting can carry the feeling before I add any dialogue.
Ada F.Picture book author
Packaging concepts are easier to discuss when I can see the box on a shelf. I describe the paper texture, the color blocks and the viewing angle, then use the image to decide which direction deserves a physical mockup.
Callum FraserPackaging designer
My recipe handout needed a close view of a dumpling fold. I wrote the image prompt around the hand position and the edge of the wrapper. I check those details carefully before using an illustration to teach a real technique.
Esme LaurentCooking instructor
Same reading nook, different mood. I explore warm wood, muted green and softer evening light with a reference image before choosing materials for the room.
Felix BraunInterior stylist
For a slide about fragile systems, I wanted a bridge made from folded paper. GPT Image 2 helps me develop a visual metaphor around the point I am making, instead of settling for an unrelated stock photo.
Imogen ClarkePresentation designer
I plan my social post cover around where the text will go. Asking for the main object off to one side gives me a composition to work with; I add the final wording and check the crop in my publishing layout.
Rhea ShahCommunity editor

Questions before you start

Can I start with a photograph

Yes. Upload a reference image and describe the change you want. Explain which details should remain, such as the subject or product shape, and which parts can change. Check the result rather than assuming every detail will be preserved.

Which image files can I add here

This page accepts up to five JPEG, PNG or WebP reference images, with a maximum size of 20 MB each. Use a clear image when the shape, appearance or layout of an existing subject should guide the task.

Can I put words inside an image

Include the exact wording and describe its position and reading order. Keep the message short when readability matters. Check spelling, punctuation and small text in the final image before using it.

How do I get a more consistent style

Keep the important parts of the brief stable, including palette, materials, lighting and composition. Use a reference where useful and change one part of the direction at a time. Review each result for differences.

Can I make an image for a specific layout

GPT Image 2 supports custom pixel dimensions within its model limits. The size options exposed in a creation task depend on that workflow. Include the intended placement and composition in your brief, and check the available output options for your layout.

How does image generation use credits

Image generation uses your account credits. A free Skill does not make the underlying model usage free. The pricing section shows current plans and credit allowances; the credits used for an image depend on the model and generation settings.