Images and video

Image and video generation live in the same chat as everything else. Describe what you need, pick a generation model, the result appears inline.

Generating an image

Pick an image model from the model picker, or click the generation button in the prompt box, and describe the image. Available model families include DALL-E and gpt-image, Imagen, and Stable Diffusion variants; the picker shows what your plan includes.

Refine with follow-ups: "same scene at night", "make it wider", "remove the text". Each generation appears in the conversation, and you can download any of them full-size.

Parameters

Generation models expose their own parameters: size and aspect ratio, quality tier, number of variants, style controls where the model supports them. A parameter panel appears when you select a generation model, showing exactly what that model accepts: the options come from the model itself, so the panel never offers settings the model would ignore.

Video

Video generation works the same way where a video-capable model is available: describe the clip, set duration and format in the parameter panel, generate. Video jobs take longer than images; the result lands in the chat when ready.

Cost

Images and video are billed per generated unit, not per token. The price depends on the model, resolution, and (for video) length, and is shown with the result. Generation is also available via the API under /v1/images/generations: see Endpoints.