The Image editor makes images by conversation. You tell it what to create or change, and the results appear in a gallery on your board. It can use images, videos, websites and notes on your board as reference.
How to start
Pick whichever fits:
From an image or video on your board: hover it and click Edit image next to its + dot. On a YouTube video the button says Create thumbnail. You can also select the card and press E.
From an image card: select it and click AI Edit in the toolbar. On a generated image you can also hover it and click Tweak. The editor opens with that image and its original prompt.
From a board chat: click Tools and choose Create Image.
Make an image
Type what you want in the editor, for example "a cat swimming on the beach". Type @ to point at a specific image.
Add references: connect cards from your board, or click Add images. You can use up to 14 reference images.
Choose settings in the toolbar (see below).
Click Send. A gallery appears on the board with your images.
Ask questions or follow-ups in the same editor, like "make it vertical" or "change the background to blue". Poppy replies in the editor. If you send more messages while it's working, they wait their turn.
Settings in the toolbar
Model: GPT Image 2.5 Flare (default), GPT Image 2.5 Sunburst, NanoBanana 2.1 or NanoBanana Pro. Turn on Use multiple models to compare up to 3 at once.
Aspect ratio: Auto (follows your reference), Square, Horizontal (16:9, 4:3) or Vertical (9:16, 4:5).
Quality: Low, Medium or High.
Variants: how many versions to make of each image. You can also type it, like "make 4".
History: your past requests. Search them and click one to reuse its prompt and settings.
New: opens a second editor next to this one.
Limits: 10 images per message, 20 per request.
Galleries and versions
Every run lands in a gallery card connected to the editor. Versions of one image stay together in one gallery. If you ask for several different images, each gets its own gallery.
Click an image to focus it. Click Show all variants to go back to the grid.
Hover an image for Open (large preview), Add to board (saves it as its own image card) and the download menu (Download image or Copy image).
Click Edit on a variant to start a new round from that image.
The Image editor chip
The editor stays open after it makes images, so you can keep going. To tidy your board:
Select the editor and click Minimize in its toolbar.
It folds into a chip labelled Image editor with the number of runs.
Double-click the chip, or select it and click Expand, to open it again.
The chip stays connected to its galleries.
Images from your Vaults
When you place an image from a Vault on your board, the card now fits the image's own shape, so portrait images are no longer cropped.
Credits
Each image uses credits. The cost depends on the model and quality, and it is multiplied by the number of variants and models. Per image, about:
GPT Image 2.5 Flare / Sunburst: Low 0.6, Medium 1.4, High 5.3, plus about 1 per reference image
NanoBanana 2.1: 4 (1K), 6 (2K)
NanoBanana Pro: 15 (1K or 2K), 30 (4K)
Examples
Ad recreation: connect a competitor's ad, your logo and your website, then ask for the same style with your branding.
YouTube thumbnails: click Create thumbnail on a YouTube video and describe what to change.
Brand images: connect your website and brand examples, then ask for on-brand graphics. You can also pick a brand in your chat. See Brand Voices.
To make a whole set of slides at once, see Carousels.
If an image fails
"Choose an image model below, then try again": pick a model in the toolbar.
"That request couldn't get through": open History, click your message and send it again. No images were started.
"Your prompt was blocked by the image model's content filter": reword it or switch models.
Other errors: see Chat Errors.
Tips
Start with a reference. Connect images you want to match.
Name the format. Say "for YouTube" or "for Instagram" or set the aspect ratio.
Iterate. Ask for specific changes in the same editor instead of starting over.