Agent

The Agent helps you create images and videos through a natural-language conversation. Describe what you want in your own words, add images when helpful, and continue with new instructions to refine a result or create something completely new.

Open the Agent

From Home or the app sidebar

Select Agent in the app sidebar, or enter a prompt from Home. You can start a new conversation or reopen an earlier one from the conversation history.

From Gallery

Agent is also available at the bottom of Gallery. Enter a request in the Agent message box, or add an image from Gallery when you want to use it in your request.

How it works

1. Describe the result: Name the subject, product, setting, or change you want in plain language.
2. Add images: Upload images, choose them from Gallery or Elements, or reference an image from the current conversation.
3. Generate: Agent chooses the appropriate image or video process and creates the result.
4. Continue the conversation: Ask for a change, refer to an earlier image, or create a variation. Start a new conversation when you begin an unrelated workflow.

Generated images and videos appear in the conversation and are also saved to Gallery.

Several results at once

When one request asks for several independent results, such as different angles, colorways, or backgrounds, Agent starts them at the same time instead of one after another. The results appear together in a grid, and each one shows as soon as it is ready. When the set is done, select Download all to export every result.

For example:

Place the model in this photo in three settings: a beach at sunset, a busy city street, and a minimal studio.

• How many run at once: up to your plan's simultaneous generations limit. When a request needs more, the extra generations wait and start as earlier ones finish. See Limitations.
• Steps that build on a result: editing, upscaling, or animating a new image starts after that image is ready. When Agent creates a new model, more shots of that model start after the first one is ready.
• Credits: each result costs the same as it would on its own. See Credit Costs.

What you can do

Limitations

• Agent cannot use uploaded videos. It can upscale only videos it created in the same conversation.
• Images cannot be fetched from external URLs.
• Account and billing changes are not supported.
• Your plan limits how many generations can run at the same time, shared across your organization. Extra generations wait and start as earlier ones finish.

What's Next?

• Learn how to create and edit images.
• Use image references to identify a specific input or earlier result.
• Learn how to create videos, including how to use a start image and an end image.
• Upscale an image to increase its resolution.
• Create transparent images or remove a background.
• Choose your generation settings for aspect ratio, resolution, speed, and quality.