Image Generation
Overview
With the chat's image generation mode, you can have the AI generate images from text descriptions. You can choose a style (art style) and aspect ratio, and also provide reference images for generation.
Note: Image generation is available only if enabled in your organization's settings (an administrator configures the API key and the models to use). In organizations where it is disabled, it does not appear in the menu.
Starting Image Generation Mode
- Click the "+" button (tool menu) to the left of the chat input box and select "Image Generation"
- On the new chat screen, you can also switch via the "Generate image" shortcut chip
- An "Image Generation" chip appears below the input box, and image generation mode is enabled
- Write a description of the image you want in the input box ("Describe the image…") and send it
To exit the mode, click the "x" on the chip ("Switch back to Standard Chat").
Choosing a Style (Art Style)
Click the "Image Generation" chip itself to open the "Image style" selection panel.
Styles provided by default:
| Style | Result |
|---|---|
| Photo | Photorealistic |
| Illustration | Hand-drawn illustration |
| Anime | Cel-shaded anime |
| Watercolor | Soft watercolor painting |
| Pixel | Pixel art / retro game |
| 3D | 3D CG render |
| Line art | Black & white line art |
| Flat / Logo | Flat logo / icon |
- When you select a style, the chip label changes to include the style name, e.g. "Image Generation · Anime"
- The selected style is automatically applied to subsequent generations
The styles displayed may vary depending on your organization's settings (administrators can disable or adjust them individually).
Choosing a Size (Aspect Ratio)
Click the "Size" chip next to the style chip to open the "Select image aspect ratio" panel.
| Option | Ratio |
|---|---|
| Auto | Automatically selected based on the content |
| Square | 1:1 |
| Portrait | 3:4 |
| Story | 9:16 |
| Landscape | 4:3 |
| Widescreen | 16:9 |
Using Reference Images (image-to-image)
You can generate images based on images you already have.
- Attach images via "Add file or photo" in the tool menu (pasting from the clipboard also works; up to 8 images)
- With image generation mode still active, write a description and send it
- The image is generated using the attached images as references
Note: To use reference images, a model that supports image-to-image must be allowed in your organization. If no compatible model is available, an error message is displayed.
During Generation and Viewing Results
When you send the message, a generation card appears in the conversation.
- While generating, the elapsed time is shown as "Generating... Xs elapsed". Depending on the model and content, it may take a few minutes
- The input box is temporarily disabled during generation (you can abort with "Cancel")
- When complete, the image and the time taken are displayed, and you can save it with the "Download" button
- If generation fails, the error details are displayed and you can try again with "Retry"
The name of the model used is shown on the card (the model is selected automatically based on your description and settings).
Where generated images are stored
Generated images also appear under "Generated" in the file list of the right panel. From there you can download or delete them.