Generated images

Every image the Co-Pilot generates is saved as an artifact, and every revision is kept as its own version. Ask for a change in the chat and a new version is added — earlier versions stay available.

Where images live

A new image appears as a card in the chat. Click Open in Canvas to open it on the canvas next to the conversation. It is also saved among your artifacts. Images without a title carry the default name Generated Image.

From a whiteboard photo to a graphic recording

After a workshop the ideas are often on sticky notes or a whiteboard. Take a photo, and let the Co-Pilot sort the ideas and draw them.

1. In the Co-Pilot, click the paper clip below the message box to attach the photo, or paste it into the box. PNG, JPEG and WebP pictures of up to 20 MB work, up to 16 per message.
2. Click the plus, then Create, then Image, and say what the new picture should do with the ideas — for example, group them into themes with one heading per theme. The Co-Pilot reads your photo first, including the text on every note.
3. The new picture appears as a card in the chat. Click Open in Canvas to see it next to the conversation. The ideas are sorted into themes, not copied from the board.
4. In the same chat, ask for a graphic recording of the topics and the positions within them. The Co-Pilot draws each topic with its positions, and also uses the discussion's transcript. Each change is saved as a new version.

The image model may run outside the EU. By default it gets only the Co-Pilot's written description, not your photo. If the Co-Pilot wants to use the photo itself, the app asks you first: Send outside the EU or Don't send. If you don't send it, the picture is still made, from the description, and your photo stays in the EU.

A generated picture can misspell or leave out a note. Check it against your photo before you share it.

You don't need a photo: posters, infographics and graphic recordings can also be made straight from a discussion's transcript. These are pictures made by image generation, not the interactive Graphic.

Controls at a glance

Editing an image

There are two ways to change a generated image. Both save the result as a new version — nothing is overwritten.

Describe the change in the chat

Ask the Co-Pilot directly in the conversation — "Make the sky evening light", "Remove the text in the corner". It regenerates the image and adds the result as a new version, with the earlier versions still in place. This route works best for changes that affect the whole image.

Brush editing on the canvas

To change only one part of the image, open it on the canvas and click Edit. A toolbar appears with the hint "Paint the area you want to edit, then describe the change below."

1. Paint over the region you want to change. Adjust the stroke width with the Brush size slider.
2. To correct the marking, switch to the Eraser and rub out the excess, click Undo to take back the last stroke, or click Clear mask to start the marking over.
3. In the text field at the bottom — "Describe the desired edit..." — say what should appear in the marked area, then click Apply edit or press Enter.

While "Applying edit..." is shown, the Co-Pilot changes only the marked region and leaves the rest of the image untouched. The result is saved as a new version and the view jumps to it automatically. Clicking View leaves edit mode at any time without applying anything.

How versions behave

• Each time the Co-Pilot regenerates the image — a different style, changed wording, added detail, a brush edit — the result becomes a new version.
• The latest version is shown by default; use the arrow buttons to compare it with earlier attempts.
• Nothing is overwritten: older versions remain in place, so you can always step back to the one you preferred and ask the Co-Pilot to continue from there.

Related articles

• The canvas: where your documents live
• Dashboards