The Image Agent is a conversational way to create and refine images. You start with a prompt and keep the conversation going until the image is exactly what you want.
What it does
You describe the image you want and the Image Agent generates it. From there, you keep chatting to make changes, like "make it bigger," "change the color," or "add a shadow." Each edit is saved as its own version, so you always have a history of how the image evolved and can pick up from any point.
How to use it
- Open the Image Agent and type a prompt describing the image you want.
- Review the result. If it's close but not perfect, type a follow-up describing the change.
- Keep refining with short, specific instructions until the image looks right.
- Use the version history to revisit an earlier image whenever you want.
- Branch from any saved version to take the image in a new direction without losing your current one.
Tips
- Make one change per message. Small, focused edits land more reliably than long, combined requests.
- Be specific about what to change and what to keep, so the agent doesn't alter the rest of the image.
- Branch from an earlier version when you want to explore an alternative instead of overwriting your progress.
- Reuse a generated image in other image tools to stage it, expand it, or add shadows.
Ready to start a conversation? Get started or learn how to place a product in a scene with Product Staging.