AI Image
AI Image uses image. Select a model from generation_models[], then use an aspect ratio supported by that model.
Nano Banana 2
{ "type": "image", "prompt": "A clean course cover about binary search in a dark vector style", "aspect_ratio": "16:9", "output_quality": "2K", "image_generation": { "provider": "gemini", "model": "nano-banana-2" }}
Current ratios are 16:9 / 9:16 / 1:1; always use the catalog as the source of truth.
GPT Image 2
{ "type": "image", "prompt": "A precise binary search poster for a developer course", "aspect_ratio": "3:2", "output_quality": "2K", "image_generation": { "provider": "azure_openai", "model": "gpt-image-2", "quality": "medium" }}
Current GPT Image 2 ratios are 3:2 / 2:3 / 1:1. Native quality uses low / medium / high and is separate from the top-level final output specification output_quality.
Use the top-level aspect_ratio and output_quality fields shown in the Public OpenAPI schema.
Choose AI Image for a single visual result
Use AI Image for an illustration, cover, visual reference, or diagram-like composition. Describe the subject, intended audience or use, composition, and visual direction. Generated imagery should still be reviewed before it represents a factual source: exact wording, dense tables, and legally significant labels need human verification rather than a prompt-only assumption.
Match the provider and model fields
Read /catalog before composing image_generation. Select a model, then use only that model's permitted aspect ratios and quality values. In OpenAPI, the provider and model sit inside image_generation, while the requested output specification stays at the top level. Do not move the flat MCP image fields into this request or combine the ratio from one provider with another provider's model.
After POST /creations, poll the job and retrieve the finished detail to obtain the returned main_url and artifacts. If validation fails, update the catalog-derived settings or clarify the prompt before submitting a distinct retry.
Related guides
Catalogfor provider and model selectionCreations & Jobsfor the asynchronous result lifecycleAI Videofor a motion result with native audio
