OpenAI: DALL·E 3

DALL·E 3 is OpenAI’s text-to-image model. Use a written brief to create an illustration, a scene concept or a visual direction for a design. Describe the subject, environment, composition, lighting and treatment in the prompt. For example, a concept brief can ask for a delivery bicycle beside a small shop at dusk, viewed from the street, with a clear foreground subject and a restrained colour palette.

Send POST /v1/images/generations with an application/json body. The exact public model ID is openai/dall-e-3; the admitted alias dall-e-3 selects the same model card. The documented starting shape supplies model, prompt, n:1 and size:"1024x1024". n is an optional integer and defaults to 1. Send stream:false, or omit stream, for the complete JSON result. Use a nonempty text prompt and keep this request focused on one complete image brief. The offered method is image generation from text.

Make the requirements that matter to the placement explicit. An editorial illustration can name the central idea and the objects that express it. A scene concept can specify the camera position, time of day and which parts of the composition need room around them. Keep exact wording that must appear in the image distinct from descriptions of the scene, then inspect the rendered lettering and visual arrangement before using the selected result.

The generation response is JSON with a created timestamp and a data array. The documented result item includes url and may include revised_prompt, prompt_filter_results and content_filter_results. Preserve those fields with the complete response, and download the image at every returned signed Ailoot URL. Ailoot signed file links last 24 hours from issue. Keep the exact returned URL, including its query string, until the file is saved; determine the saved image format from the actual response or file bytes.

Treat the returned image file as the usable design asset and the JSON as its accompanying generation record. Review the image against the supplied brief before placing it in a page, presentation or application. If the response is incomplete or the send outcome is uncertain, preserve the original request identifier and returned information while checking the existing attempt rather than creating another image solely to recover the missing result. To retrieve saved results, use the same API key to GET /v1/media/assets?family=image and follow every next_cursor. The asset ID appears in each returned /v1/media/assets/{asset_id}/content URL and in its media-library record. Read GET /v1/media/assets/{asset_id} for each selected record and download its exact content_url. If the file request returns 404 with code link_expired, read that same asset again to obtain a refreshed link. Refreshing an asset link does not require another image-generation request.