Generating images from a description Guide

The AI Assistant can create an image directly from a written description, so you can go from an idea to a finished graphic without leaving the chat. Once an image is generated, it stays in the conversation, ready to preview, download, or reuse in a later message.

Generate an image from a description

The clearest way to start is simply to ask. Describe what you want in plain language and the assistant produces the image in its reply.

1. Open a chat with the AI Assistant.
2. In the composer, describe the image you want. Be specific about the subject, style, mood, and any colors or composition that matter to you.
3. Press Send (or Enter). The assistant generates the image and displays it in its reply.

You can also start an image request quickly by typing / in the composer and choosing the image command, then adding your description.

Tip: The more detail you give, the closer the first result tends to be. Mention the purpose (for example, a social post, a header, or an icon), the visual style, and anything that must or must not appear.

Write a strong description

A good description usually covers a few things at once. You don't need every element, but including several helps the assistant match your intent.

• Subject: what the image is of.
• Style: photographic, illustrated, flat, minimalist, hand-drawn, and so on.
• Mood and color: bright and energetic, calm and muted, a specific palette.
• Composition: close-up, wide shot, centered subject, room for text.
• Use case: a banner, a thumbnail, a slide background, a product mockup.

Refine the result

If the first image isn't quite right, you don't have to start over. Because the assistant remembers the conversation, you can ask for changes in your next message.

• Ask for a specific adjustment, such as "make the background darker" or "try a flat illustration style instead."
• Ask for a different framing or orientation, like a wider shot or a square version.
• Request a new variation on the same idea if you'd like options to choose from.

Each request builds on what came before, so you can iterate until the image matches what you had in mind.

Download and reuse the image

Once you're happy with an image, you can keep it and use it elsewhere.

1. Download it to save a copy to your device for use in other tools.
2. Reuse it later in the chat by referring back to it in a new message. For example, you can ask the assistant to include the image in a document it's creating, or to use it as a starting point for a related graphic.

Images and files shared in a conversation are collected in that chat's Files panel, so you can return to a previous image and preview or reuse it without scrolling back through the whole conversation.

Use generated images in other deliverables

Generated images don't have to stand alone. The assistant can produce polished deliverables in the same chat, so you can ask it to bring an image into a broader piece of work, such as a document or a slide, as part of a single request.

Note: When the assistant does something that reaches outside the chat, such as sending or sharing your image through a connected app, it shows an Approve card first. Nothing leaves the conversation until you confirm it.

If an image doesn't turn out as expected

• The result missed part of your intent: reply with the specific change you want rather than re-describing the whole image from scratch.
• You want a cleaner match: add detail about style, color, and composition, and name anything that should be excluded.
• You'd like to compare options: ask for a variation or an alternative style, then keep the version you prefer.