Image Generation and Processing

Image models expose one or more operations, such as text-to-image generation, reference-image generation, or image editing. Supported operations and controls belong to the model, not to its manufacturer as a whole.

Choose a model and operation

Check the selected model for:

• Supported generation or editing operations.
• Accepted input modalities and reference-image count.
• Size, aspect ratio, resolution, quality, and output-count controls.
• Supported result formats and execution modes.
• Current price and model-specific limits.

Do not send an editing or reference-image request merely because another model from the same manufacturer supports it.

Choose an endpoint

The OpenAI-compatible surface currently provides POST /v1/images/generations and POST /v1/images/edits. Other supported models can retain a documented native public request shape when their operations cannot be represented by the same schema.

Use Protocols and SDK Compatibility to configure the client, then send only operations and parameters supported by the selected model.

Inputs and results

Input images can be carried as multipart uploads, URLs, or encoded data when supported by the protocol and model. Results can be returned as image URLs or base64 data.

Treat generated URLs as temporary and copy assets you need to keep into your own storage. See Data and Result Retention.

Billing

Image requests can contain several billable meters, including input images, output count, size, resolution, quality, tokens, or other model-specific attributes. The model page shows current prices, and request logs show the final amount.