Skip to navigation

Generate

Each generate endpoint runs one model: POST /v2/image/generate/{model}. They share a request shape, so switching models usually means changing the path and the model’s own options.

Fields every generate endpoint takes

FieldWhat it does
promptWhat to generate, up to 10,000 characters. Ideogram 4.6, 4.5, 4.0 and P-Image Ideogram also accept a structured JSON prompt
num_images1 to 8 images per request
asyncReturn a generation_id straight away instead of waiting. Poll GET /v2/generations/{generation_id} for the result
webhook_urlReceive the result at your HTTPS endpoint. Every endpoint except auto takes it. See Webhooks

Most models also take seed, aspect_ratio or a size field, and their own quality or speed setting. Each endpoint’s page lists every field.

Edit with a prompt

auto, Ideogram 4.6, Ideogram 4.5, GPT Image and Nano Banana also accept source images in a multipart/form-data request. The prompt then describes the change to make. To change one region and keep the image’s exact size, use Precise Edit.

Choose a model

Choose a model compares the models’ inputs, controls and prices. Start with Ideogram 4.5, or send the request to auto and let Ideogram pick.

The same models run in the Ideogram app. See Models and Create Images in the user guide.