Skip to main content
Tips for marking areas:
  • Draw the rectangle slightly larger than the area you want to change. This gives the AI enough context to blend naturally.
  • Mark separate changes as separate regions with their own instructions.
  • Avoid overlapping rectangles. Each region should cover a distinct area.
Mark & Edit gives you precision control over image edits. You draw rectangles on the parts of an image you want to change, write instructions for each marked area, and the AI edits only those regions. Everything outside the marked areas stays untouched. This is a direct canvas operation, so you do not need to send a prompt to the agent.
Using canvas tools directly gives you more control and uses significantly fewer credits than sending a prompt to the agent. For targeted edits, always use Mark & Edit instead of describing changes to the agent in the chat.

​
How to use Mark & Edit

1

Select an image on the canvas

Open the canvas in the side panel and click the image you want to edit. The image editing toolbar appears.
2

Click Mark Edit

In the toolbar, click the Mark Edit button. The editor switches to marking mode. Your cursor becomes a crosshair.
Canvas toolbar with the Mark Edit button highlighted, cursor changed to crosshair
3

Draw rectangles on the areas you want to change

Click and drag to draw a rectangle around each area. Rectangles are numbered automatically (1, 2, 3, and so on). You can mark as many areas as you need on the same image.
Image showing two numbered rectangular regions marked — region 1 around a headline, region 2 around a background section
4

Write instructions for each marked area

For each numbered rectangle, type a description of what you want changed in that area.

Change an element's color in a specific region.

Replace text in a marked area.

Remove an element and fill with surrounding context.

Add a visual element to a region.

Replace an entire section of the image.

You can also attach a reference image for any marked region. The AI uses it as a visual guide for style, color, or content in that specific area.
5

Click Apply

The AI processes all marked regions at once and generates a new image with your changes applied. Only the marked areas are modified. Everything outside them is preserved.
Two canvas items side by side — the original image and the edited version with changes visible only in the previously marked regions
Mark & Edit creates a new canvas item. It does not overwrite the original. You always keep both versions.

​
When to use Mark & Edit

Mark & Edit is the most flexible editing tool in AI Images. Use it when:
  • You want to change a specific part of an image without affecting the rest.
  • Edit Text did not detect text you need to change (stylized fonts, low contrast, small text).
  • You want to replace, remove, or add elements in specific regions.
  • You need to make several different changes to different parts of the same image in one pass.
  • You have a reference image for the look you want in a specific area.

​
How Mark & Edit differs from other tools

ToolBest for
Mark & EditPrecise, region-specific changes. You control exactly what gets edited and what stays.
Edit TextChanging wording only. Faster when the text is cleanly detected.
Remove BackgroundRemoving the entire background. One click, no marking needed.
Replace BackgroundSwapping the entire background for a new one. One click with a prompt.
Asking the agentComplex multi-step edits, generating entirely new images, or combining multiple operations.

​
What you cannot do

  • You cannot draw freeform shapes. Marked regions are always rectangles.
  • You cannot resize or reposition a marked rectangle after drawing it. Delete it and draw a new one.
  • You cannot undo after applying. If the result is not right, use Rollback or run Mark & Edit again on the original image.
  • You cannot use Mark & Edit on a video frame. This tool works on images only.

​
Next steps

AI Canvas overview

Explore the full canvas workspace where all image editing happens.

AI Videos overview

Turn your images into videos or generate video from text.