What is AI Text to Image?
AI Text to Image is a powerful feature in Abyssale's Design Builder that allows you to generate custom images directly from text descriptions (prompts). Instead of searching for stock photos or creating graphics from scratch, simply describe what you want, and our AI will create it for you.
This feature is perfect for:
Creating unique visuals for your marketing campaigns
Generating custom illustrations for your designs
Exploring creative concepts quickly
Producing on-brand imagery without professional design skills
How to Generate Your First AI Image
Select any image layer on your canvas. An AI toolbar automatically appears at the bottom of the canvas, giving you quick access to Abyssale's AI tools for that layer:
Text to image
Click Text to image to start generating a new visual for this layer.
Write your prompt
The toolbar transforms into the AI Text to Image prompt bar. This is where you'll build your generation:
Prompt field : describe what you want to see, in as much detail as possible.
Image refs : attach up to 4 reference images to guide the generation (style, composition, subject, etc.). On the right, you'll see a thumbnail of the image currently in your layer
Note that this existing image is not used as a reference automatically; click it if you want to include it as one of your references.
Model : choose which AI model generates your image (see below).
Ratio / Dimension and Quality : depending on the model selected, additional options appear to control the output format and quality level.
Variations : use the layers selector to generate 1 or 2 variations at once from the same prompt.
Choosing a model
Click the model dropdown to browse all available AI models. Each one is listed with its relative AI credit cost and, where relevant, a badge such as Top Pick or New:
Model |
| Default |
| Default |
|
|
|
|
|
|
|
|
|
|
|
|
| Not supported | — |
|
|
|
|
|
|
|
| Not supported | — |
|
|
| Not supported | — |
|
|
| Not supported | — |
|
|
|
|
|
|
|
|
|
|
|
|
| Not supported | — |
|
|
|
|
|
|
|
|
|
|
|
|
| Not supported | — |
|
|
| Not supported | — |
|
|
|
|
|
Some models also support image references directly, marked with an Image ref tag in the list.
Improving your prompt
Above the prompt field, Abyssale suggests contextual tags (for example Enhance lighting, Improve text, Refine branches, Boost color) to help you add useful detail with a single click.
You can also click Enhance to have the AI rewrite your prompt for clarity and detail, without changing its substance or creative direction.
Check your credit cost and generate
The number of AI credits your generation will cost is calculated automatically based on the model, quality, dimension, number of image references, and number of variations you've selected. This cost is displayed right next to the arrow (generate) button, so you always know before you launch.
Once you're ready, click the arrow button to launch the generation.
Note: While your image is generating, you can't modify anything else in the editor. If you try to navigate away during this time, a warning will appear:
"You'll lose this AI generation if you leave. Credits already used won't be refunded."
You can choose Keep waiting to let the generation finish, or Proceed anyway if you're willing to lose it.
Working with Your Generated Images
Selecting a variation
Once generation completes, you'll see thumbnails for each variation produced. Click on the one you prefer to select it, a green outline shows your current selection.
From here, you have two options:
Apply : insert the selected image directly into your layer.
Iterate : describe your changes in the prompt field to refine the selected image further. The same Enhance button and suggestion tags are available here too, so you can keep improving the prompt with each iteration.
At this stage, you can also click Download to save the generated asset to your device if you'd like to keep a copy outside your design.
Saving your AI action
After you apply your selection, a prompt appears asking:
"Use this AI action when generating assets? It will be used automatically for future generated assets."
Choose Save this AI action if you want this same prompt and settings to automatically re-run whenever new assets are generated from this design (for example, across multiple formats or bulk generation).
Choose No thanks if this was a one-off generation or click away.
Revisiting your AI generation settings later
You don't need to reopen the AI toolbar to check or adjust a saved generation. From the Style panel on the right, with your image layer selected, open the AI generation section. From there you can review and edit the Text to Image settings (prompt, model, ratio, quality, image references) whenever you need to, without starting a new generation from scratch.
Writing Effective Prompts
The quality of your text prompt directly impacts the quality of your generated image. For detailed guidance and best practices on creating effective text-to-image prompts in Abyssale, please refer to our dedicated article:
Generate AI Images via API
You can also leverage the power of AI Text to Image directly through the Abyssale API. This lets you generate unique visuals dynamically during your automation workflows or edit an existing image with AI instead of providing a static image URL.
Note: Text to Image via the API is only available for asynchronous generation. It's also ignored if image_url or image_encoded is provided on the same image element.
How to structure the payload
On your image element, activate text_to_image and pass your settings inside text_to_image_properties:
Parameter | Type | Description | Default |
| Boolean | Activates AI image generation |
|
| Object | Generation settings | — |
| String | Description of the image to generate, or of the edit to apply in inpainting mode | Required |
| String | Model to use — see Models below |
|
| String | Aspect ratio or size of the output. Allowed values depend on the model | Model-dependent |
| String | Output quality/resolution. Only supported by some models | Model-dependent |
| Array of strings | URL(s) of the image(s) to edit. When provided, switches the generation to inpainting mode instead of pure text-to-image | — |
Text to image
Without inpaint_images, a brand new image is generated purely from your prompt:
{
"elements": {
"image_0": {
"text_to_image": true,
"text_to_image_properties": {
"prompt": "A sleek, modern glass villa in the middle of a minimalist lavender field. Sharp lines, reflection of the blue sky in the windows, wide-angle shot, architectural photography style, hyper-realistic.",
"model": "nano-banana-2",
"ratio": "16:9",
"quality": "2K"
}
}
}
}Inpainting
Pass one or more URLs in inpaint_images to have the model edit those existing images according to your prompt, instead of generating from scratch:
{
"elements": {
"image_0": {
"text_to_image": true,
"text_to_image_properties": {
"prompt": "enhance the product by adding a background decoration to highlight the product",
"model": "gpt-image-2",
"ratio": "square_hd",
"quality": "low",
"inpaint_images": [
"https://your-storage.example.com/path/to/reference-image.jpeg"
]
}
}
}
}Note: Because the image is generated in real time, banners using text_to_image may take slightly longer to generate than banners using a static image_url.
For the full, always-up-to-date reference, see the Image element properties developer documentation.
Writing Effective Prompts
The quality of your text prompt directly impacts the quality of your generated image. For detailed guidance and best practices on creating effective text-to-image prompts in Abyssale, please refer to our dedicated article:
Troubleshooting: Image Doesn't Match My Prompt
If your generated image doesn't match your expectations:
Try adding more specific details to your prompt, or use the suggested tags to fill in missing detail
Use the Enhance button to have the AI tighten your prompt without changing its direction
Try a different model, some models are better suited to photorealism, others to illustration or text rendering
Break complex scenes into simpler elements
Iterate from your current selection rather than starting over, to keep what's working while refining what isn't