Text to Image
With Text to image you create images directly from the dashboard based on a description. This is a direct input mode and therefore does not work through the automatic tool selection in chat.
Starting from the dashboard
On the dashboard, choose Text to image below the input field. The input field becomes larger, so you can write a detailed image description. The rest of the dashboard remains available, but the input gets more space.
Then write what you want to create, for example:
Create an accessible visual for public communication about a green neighborhood.
Use a clear, friendly style and little text in the image.
Settings
Via the settings button next to the input field, you can adjust the image generation.
| Setting | Explanation |
|---|---|
| Model | Choose the image model you want to use. |
| Format | Choose Automatic, Square, Landscape, or Portrait. Automatic is the default. |
| Style | Choose, for example, photorealistic, illustrative, editorial, clean design, or playful. |
| Variants | Choose how many images you want to create: 1, 2, 3, or 4. By default, 2 variants are created. |
| Color preference | Enable color preferences and add your own colors or brand colors. |
You can save these settings as your personal default. With Restore defaults you remove your saved image settings.
Automatic format
The format is set to Automatic by default. AI-Public then looks at your prompt and, in an existing chat, at relevant earlier messages. If you ask for a poster, portrait may make sense. If you ask for a presentation slide, landscape may fit better.
If you choose Square, Landscape, or Portrait yourself, that choice remains leading.
Adding inspiration images
You can add images as inspiration. The AI uses these images as visual guidance for style, color, composition, or subject. The result is still a new image.
Use inspiration, for example, when you want a new image to get the same mood or style as a previously generated image.
Editing images
You can have an existing image adjusted. Via the settings button next to the input field, choose Add image to edit and select an image. You can also choose a PDF; AI-Public then uses the first page as an image.
In edit mode, you do not choose an aspect ratio. AI-Public then does not send a manual format, so the model can preserve the dimensions and proportions of the source image.
There are two ways to indicate what should change. You can also combine them.
With a general instruction
Describe in the input field what should be adjusted. This works well for changes to the entire image.
Examples:
- "Make the background lighter and calmer."
- "Remove the text and keep the composition."
- "Change the word Certificate to Diploma."
With annotations
With annotations, you indicate directly in the image what should change. This means you do not need to describe in detail which part you mean.
- Add the image for editing and choose Open and annotate. The image opens large next to the chat.
- Click the Annotate button.
- Drag with the mouse around the area of the part you want to adjust, for example a word, a figure, or an animal.
- Write in the window what should happen to this part, for example "make it yellow", "change this to Document", or "make this twice as small", and choose Save.
- Repeat this for other parts. Each area gets its own instruction.
- Send the edit. An extra instruction in the input field is not needed.
The area does not have to be exact. If you circle a word or figure only partially, the whole part will still be adjusted.
If you want to change something about the entire image as well as the indicated parts, type that in the input field, for example "make the background yellow". That instruction is applied to the entire image after the annotations.
Result
The generated images appear directly in the chat. You can open them, download them, and reuse them as inspiration or as a source for an edit.
When downloading an image version, choose Download as image or Download as PDF (A4). The PDF places the image on an A4 page, so you can print it directly. AI-Public chooses portrait or landscape automatically, depending on the image. The text in the PDF is part of the image and therefore cannot be edited separately.
Difference from the Create Images tool
The Create Images tool is used by the AI from the regular chat when you ask for an image. Text to image is more direct: you consciously choose image mode and set model, format, style, colors, and reference images yourself.
Editing subsequent image versions
In image mode, OpenAI and Gemini automatically edit the previous image version on a follow-up question. Select an older version to edit it; above the input field you can see which version is used as the source. A manually selected edit image takes precedence. After a successful generation, the new image becomes the source for the next edit. Flux/BFL and Stability AI do not edit images and do not receive an automatic source image. Start a new chat if you want to create a separate new image.
