Docs
Text to image
Create images from prompts and adapt settings to the selected model.
Updated 2026-07-11
Text to image
Text-to-image starts from a written description. Select a model, then use only the aspect ratio, output count, style, negative prompt, seed, or quality controls that the panel exposes for that model.
A useful prompt structure
Describe the medium, subject, action or pose, environment, composition, light, and color direction. Put required details before decorative adjectives.
Editorial product photograph of a cobalt glass bottle on pale limestone, centered composition, soft side light, warm gray background, shallow depth of field, label area clean and readable.
Iterate deliberately
Generate a small set first. When a result is close, change one variable at a time. Reference workflows are usually better than text alone when exact composition or identity matters.
Image models can produce incorrect text, hands, logos, reflections, or visual facts. Review at full size and verify third-party rights before publishing.
