oc_prompt, then the selected rendering backend creates the image. The request returns an asynchronous img_uuid for polling or callback handling.oc_prompt, then the selected rendering backend creates the image. The request returns an asynchronous img_uuid for polling or callback handling.
Supply the full text-stage prompt and define the JSON fields your application needs. The response must contain oc_prompt, while fields such as role name, appearance, personality, or backstory can pass through with the generated character data.

Select seedream or gpt-image-2 for the image stage, and optionally set a supported ratio. A request-level ratio overrides oc_ratio from the generated JSON, keeping layout decisions under application control.

oc_prompt. Add any character fields your application needs.model: oc-maker, your prompt, and a required image_render_model. Optionally set the output ratio, text model, system prompt, or callback URL.data.img_uuid with your own task record. Poll GET /api/v1/jobs/detail or process callbacks until terminal state, then read the structured object from the JSON entry and the asset URL from the image entry.