GPT-Image 2.5 image generator and editor
- Type
- Image model
- Made by
- OpenAI
- Input
- Text, plus images to edit or follow
- Output
- Image (JPEG, PNG, WebP)
- Options
- Quality: Low, Medium, High, Extra High, Max · Speed: Fast, Slow · up to 4 per run · 13 aspect ratios · Resolution: 1K, 2K, 4K
- Price
- From 25 credits
- Plan
- All plans, including Free
GPT-Image 2.5 is OpenAI's image model for both making and editing pictures. Describe what you want, or attach an image and say what should change, and it returns up to four images per run at 1K, 2K or 4K. It is a strong fit for work where details matter: exact edits, readable text, and people or products that should still look like themselves.
What you can do with it
- Text to image. Describe a scene, a layout or a design and get finished images.
- Edit an image. Attach a photo and describe the change: swap the background, change the outfit, fix the lighting.
- Follow references. Attach images of a person, product or place and describe a new picture built around them.
- Put words in the picture. Posters, packaging, slides, thumbnails and signs, with the text written into the image.
How to use it
- Write your prompt in the composer.
- Attach images if you want to edit one or have the result follow it.
- Set the chips: quality, resolution, aspect ratio and number of images.
- Check the price shown in the composer, then generate.
Writing prompts for GPT-Image 2.5
- Quote the exact text. Put the words you want in quotation marks and say where they go: "a headline at the top that reads 'Open Late'".
- For edits, say what stays. "Keep her face, pose and the window light; change the sweater to dark green" works better than only naming the change.
- Describe the medium. Photo, flat illustration, watercolor, 3D render, product shot on seamless paper.
- Give layout cues for designs: where the subject sits, how much empty space to leave, what goes in each corner.
Example prompts:
- A minimalist coffee shop poster, cream background, a single paper cup in the center, the words 'Slow Mornings' in a tall serif above it.
- Using the attached product photo, place the bottle on a wet slate surface at dusk, with soft rim light and a shallow depth of field.
Choosing settings
- Quality: Low, Medium, High (default), Extra High or Max. Higher quality costs more per image.
- Speed: Fast (default) or Slow.
- Images: 1 to 4 per run, default 2.
- Aspect ratio: 1:1, 3:4, 4:3, 2:3, 3:2, 4:5, 5:4, 9:16, 16:9, 1:2, 2:1, 9:21 and 21:9.
- Resolution: 1K (default), 2K or 4K.
- Output format: JPEG (default), PNG or WebP, in the advanced settings panel.
Credits per image are the same for new images and edits:
| Quality | 1K | 2K | 4K |
|---|---|---|---|
| Low | 25 | 50 | 75 |
| Medium | 60 | 100 | 175 |
| High | 225 | 375 | 675 |
| Extra High | 400 | 675 | 1,200 |
| Max | 900 | 1,500 | 2,500 |
Each attached reference image adds 38 credits. Two High images at 1K with one reference image come to 488 credits. A practical approach is to explore at Low or Medium, then re-run the version you like at a higher quality and resolution. The composer always shows the exact price before you generate.
Apps built on GPT-Image 2.5
Several ready-made apps use GPT-Image 2.5 with a preset prompt and options: Studio Headshot, Studio Portrait, Thumbnail Maker, Director's Storyboard and Web Designer.