Krea 3
Krea 3 is Krea's fast, edit-first image model: describe a change in plain language and it applies it in seconds while keeping products, logos and characters recognisable, or leave out the images and it generates from text. Combine up to 8 images, switch to high thinking to plan multi-step edits, and use the experimental regional edit to change one detail while the rest of the picture stays untouched.
Input
Describe the image to generate, or the edit to make: say what should change and what should stay.
Optional input images (up to 8) to edit or combine; refer to each by what it shows. Larger images cost more.
0 / 8
How much the model plans first. high handles complex, multi-step edits more carefully but is slower and costs more.
Output resolution. 2K costs more than 1K.
Output aspect ratio. auto picks the framing from the input images and the prompt.
Experimental: change only the parts that need it and keep the rest untouched. Needs an input image.
What to avoid in the output.
Random seed. The prompt is expanded before generation, so a seed doesn't guarantee identical results.
Output
Run the model to see output here.
Pricing
Features
API Examples
curl -X POST \
/api/v1/model/krea/krea-3 \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"input": {}
}'