Image Generation

Image Generation lets an Agent create images from a written prompt. Where supported, it can use reference images to maintain a visual direction and produce several variations for review. The available quality and aspect-ratio choices appear in the image tool and may vary by model or workspace.

Enable the capability in the Agent builder’s Tools area. A useful prompt describes the subject, composition, style, palette, output dimensions or aspect ratio, and any details that must remain consistent. When using a reference image, state which elements to preserve and what may change. Do not include private or copyrighted source material unless the user has permission to use it.

Review generated images for text accuracy, unwanted elements, visual consistency, and suitability for the intended use. For a series, use the same reference and describe the repeated style explicitly. Generation produces a draft visual; it does not verify brand, factual, or licensing requirements. Check each output before publication.

Output controls

The generation interface may offer quality sizes such as 1K, 2K, and 4K, and aspect ratios including 1:1, 3:2, 2:3, 4:3, 3:4, 4:5, 5:4, 16:9, 9:16, and 21:9. Use the choices shown for the selected model. Conversation previews may be compressed; download the full-resolution artifact to inspect detail.

A generation can use prior outputs or uploaded reference images for visual continuity, and several related images may be generated in parallel. Ask for consistency explicitly, and distinguish what must stay the same from what may vary. Generated images may include provenance watermarking and are subject to content filters; unsupported requests may be declined.