Cheaper Gen.
OpenAIPrompt guide

GPT Image 2.5 Prompt Guide

How to write GPT Image 2.5 prompts: a structure that works, 5 example prompts for text to image and image edit, the settings that change the result and how to reference your own media. Copy any prompt into the Studio to try it.

Example prompts
5
2 tasks
Size
4K
1K · 2K · 4K
Aspect ratios
11
1:1 · 3:2 · 2:3 · 4:3
Per try from
$0.0066/image
1K · low
Try in Studio Call the API

Structure

A GPT Image 2.5 prompt, piece by piece

Write in this order and keep each part concrete. Physical detail works better than mood words.

  1. 01

    Subject

    The main subject and its key details: material, colour, expression, product features.

  2. 02

    Composition

    Framing and angle: close-up, flat lay, three-quarter view, rule of thirds, negative space.

  3. 03

    Setting and light

    Background, environment and lighting: studio softbox, golden hour, neon.

  4. 04

    Style

    Photo, illustration, 3D render or painting — plus lens or medium details.

  5. 05

    Edit instruction

    For edits, say what to change and what must stay the same: “replace the background, keep the product and its shadow”.

Examples

GPT Image 2.5 prompt examples

Prompts written for this model, grouped by task. Open one in the Studio, then change one part at a time.

  • Text to Image3:4
    Packaging design for an oat-milk carton called “OATLY GOOD”, pastel illustration, nutrition panel with legible text
    Try in Studio
  • Text to Image9:16
    A UI mockup of a meditation app home screen, soft gradients, three cards with short labels
    Try in Studio
  • Image Edit4:5
    Replace the background with a sunlit Mediterranean terrace; keep the bottle and label exactly the same
    Try in Studio
  • Text to Image
    A minimalist poster for a jazz night called “BLUE HOUR”, saxophone silhouette, deep blue and brass, clean typography
    Try in Studio
  • Image Edit
    Keep the product and its label exactly; replace the background with a soft off-white studio sweep and add a natural shadow
    Try in Studio

By task

Prompting each GPT Image 2.5 task

Text to Image

Lead with the subject, then composition, setting, light and style. Quote any text that must appear in the image.

Image Edit

Describe the change, not the whole image. Name what must stay untouched so the model keeps it.

Referencing your own media

GPT Image 2.5 reference inputs and how to mention them
InputIn the promptLimits and rules
Input images
Image Edit · up to 16
see descriptionUp to 16 images to edit or combine. Refer to them as "image 1", "image 2"… in the prompt.

Settings

Settings that change the result

Set these next to the prompt — in the Studio or as API parameters.

GPT Image 2.5 generation settings
SettingOptionsEffect
Aspect ratio
aspect_ratio
1:1 · 3:2 · 2:3 · 4:3 · 3:4 · 5:4 · 4:5 · 16:9 · 9:16 · 21:9 · 9:21
default 1:1
Output image shape.
Resolution
resolution
1K · 2K · 4K
default 1K
Output size tier: 1K (≈1 MP), 2K (≈4 MP) or 4K (≈8 MP, experimental).
Price: 1K $0.0066/image · 2K $0.0118/image
Quality
quality
Low · Medium · High · X-High · Max
default medium
Rendering effort. Higher quality uses more output tokens and costs more; xhigh and max are for final assets.
Price: Low $0.0066/image · Medium $0.013/image · High $0.048/image

FAQ

GPT Image 2.5 prompting questions

There is no separate negative prompt for GPT Image 2.5. Say what you want instead of what you don't, and keep unwanted elements out of the description.

Image Edit: Up to 16 images to edit or combine. Refer to them as "image 1", "image 2"… in the prompt.

GPT Image 2.5 has no seed parameter, so results vary between runs. Generate a few variations and keep the best.

Open GPT Image 2.5 in the Studio and paste any prompt from this page. Each run shows its exact price before you generate, and failed generations are refunded.