Create images with AI
Create with AI in the media library makes pictures from a description and adds them to the library you have open. There are two kinds:
- Illustration or icon: a picture drawn as an SVG, sharp at any size. Choose an Illustration (a spot illustration or small scene), an Icon (one color, or two tones), a Pattern (a seamless background that tiles) or a Logo mark (a simple symbol without lettering). Available wherever Aglyn AI is.
- Photo: a picture made by Google's image models. Offered only where photo creation is turned on for the deployment your workspace runs on; where it is not, the window shows only Illustration or icon.
Each picture is stored the way a file you upload is stored: in the open folder, counted toward your storage, served from the same addresses, and with alt text you can edit.
Make a picture
- Open Media, on a site or for the organization, and open the folder the pictures should land in.
- Choose Create with AI, beside Upload media. An empty library offers it beside its own Upload media button too.
- Where photos are on, choose Photo or Illustration or icon. For an illustration, choose its Kind.
- Describe the picture: what is in it, the setting, and the style you want.
- Choose a shape: Square 1:1, Landscape 4:3, Portrait 3:4, Wide 16:9 or Tall 9:16, and how many pictures to make, from one to four.
- For an illustration, choose its colors: Use my site's theme colors, which reads your site's theme, or Choose colors to pick up to six. The organization's library belongs to no single site, so there you choose the colors.
- The window shows what the pictures will cost in credits before you start. Choose Create.
The pictures appear in the library a few seconds later, selected, so you can move, tag or open them at once. Nothing is placed on a page: a picture is used only when you put it in an Image element, a gallery or anywhere else a media field asks for one.
What each picture gets
- Alt text. A photo's is your description; an illustration's is a sentence the AI writes about what it drew. Edit it in the picture's details like any other alt text.
- A file name made from the first few words of your description, starting
ai-and numbered within the set:.svgfor an illustration, the picture's own type for a photo. - A record of how it was made: the model, your description, the mode, the kind of illustration and the shape, kept with the file.
- For a photo, an invisible watermark. Google marks every picture its image models make with its SynthID watermark, so the picture can later be identified as generated. You cannot see it, and it does not change how the picture looks.
Credits
Pictures are metered in AI credits from the workspace's pool. The window shows an estimate before you create anything:
| Mode | About, per picture | What it is made of |
|---|---|---|
| Illustration or icon | 18 credits | The words the AI reads and writes to draw the SVG, at the same rates as other AI text |
| Photo | 108 credits | The picture, plus the description and the thinking the image model does before it draws |
What is charged is what was actually spent, so a picture can cost a little more or less than its estimate, and an illustration that needed a second attempt costs about twice its estimate. You are charged only for pictures that reach your library:
- a picture the safety filter holds back is not charged;
- an illustration that does not come out right is not charged. The window says This one's on us — you weren't charged.
- a picture your library could not store, because your storage is full for example, is not charged;
- when the AI service fails, nothing is charged.
A Free workspace spends from its monthly allowance and is told before it starts when the pictures it asked for would cost more than it has left. A paid workspace past its included band keeps working at its plan's overage rate unless the workspace stops AI at the band, the same as every other AI request.
Safe by construction
An illustration is checked before it is stored. It has to be one SVG in the shape you chose, no larger than 200 KB, and self-contained: no scripts, no links, no embedded images or web pages, and nothing loaded from anywhere else, such as a font. An illustration that fails the check is drawn again once with the problem pointed out; if the second one fails too, it is not stored and not charged. The media library then removes anything unsafe from every SVG it stores, as it does for an SVG you upload.
The AI declines to draw a real company's or organization's logo, a trademark or a recognizable brand symbol, a real, named person, and anything sexual, hateful, violent or otherwise unsafe; the window says why, and a declined request on a Free workspace costs nothing. A Logo mark is always a symbol without lettering.
Photos are checked by Google's safety filters, set to block content they rate as a medium risk or higher, and Google's own policies refuse some content whatever is asked, such as photorealistic depictions of children or celebrities that its policies do not allow. A description that is declined is answered with a message and no picture is charged.
What is sent
- Illustration or icon: your description, the kind and shape of picture, and your site's theme colors or the colors you chose are sent to the AI service every other Aglyn AI feature uses.
- Photo: your description and the shape you chose are sent to Google's image models on Vertex AI.
Nothing else is sent: no other content of your site, no file from your library, and no name, email address or account identifier. The pictures come back to Aglyn and are stored in your library; your description is kept with each picture as part of its record of how it was made.
Who can use it
Creating images needs the Generate with AI permission and a plan that includes AI generation — see who can use Aglyn AI. It also needs permission to upload to the library you have open, and on a site whose AI is switched off it is not offered.