GPT Image 2 Image Generator

Generate publish-ready visuals from prompts and images

Use GPT Image 2 for posters, commerce images, infographics, brand touchpoints, and social covers. Start from a prompt alone, or upload reference images for image-to-image edits and recompositions.

GPT Image 2Current price: 6 creditsLow, medium, and high quality are priced separately
Prompt tips
  • Say what to preserve or change from the reference
  • Specify ratio, material, lighting, and hierarchy
  • For text, state the language and exact short labels
Reference images

Upload products, people, sketches, or style references. Up to 4 images. Leave empty for text-to-image.

JPG, PNG, or WebP. 8MB max each.
Aspect ratio
Image quality
Current price: 6 credits

Result

The generated image will appear here.
GPT Image 2

Choose the right GPT Image 2 workflow

Start by matching the input you have with the asset you need. That is more reliable than rewriting one long prompt repeatedly.

Prompt-only

Start from text when no fixed asset exists

Best for poster concepts, commerce hero images, infographics, brand touchpoints, and social covers. Specify the subject, format, hierarchy, text language, and use case.

Reference images

Upload references when product, person, or sketch details matter

Best when you need to preserve product shape, facial identity, color direction, or composition. Upload up to four reference images, then say what should stay, change, or be recomposed.

Quality tiers

Use quality tiers to control draft cost

Low quality uses 3 credits for direction checks, medium uses 6 credits for regular assets, and high uses 10 credits for final candidates.

Recommended workflow

  1. 1Pick the output format, such as 4:5 social creative, 9:16 story cover, or 16:9 landscape asset.
  2. 2Describe the subject, layout, material, lighting, text language, and elements that must not change.
  3. 3Validate direction with low or medium quality before moving the strongest candidate to high quality.
  4. 4Use the prompt library or examples page when you need a runnable starting point.

Useful boundaries

  • Reference images guide the output, but exact text, logos, and small details may still need review.
  • For real people, brands, or commercial use, confirm permissions, likeness rights, and source ownership first.
  • If the goal is motion, generate a stable hero image first, then continue in the image-to-video workflow.
Next actions
Browse image prompt templatesSee real examplesRead the GPT Image 2 model pageAnimate an image