Creating Images with Generative AI

AI Wizard Interface

To open the AI Wizard, click on the button in the main window. The AI Wizard lets you set up the task for the AI and start rendering.

1 - Number of remaining render coins and offers the option to purchase more.

2 - Prompt is a textual description of the image that you want to create. Parts of the prompt should be separated with commas. Only the English language is supported. Example: stones on white background, glossy, top down view.

3 - The Use Image as Input option allows the AI to use the current design as the input image.

4 - Specifies whether to use colors or only shapes from the current design.

5 - A preview of the input image.

6 - The AI Effect Amount slider determines how much of generative AI will be used in the resulting image. At the zero value, the image will be identical to the input image.

7 - The Keep Original Shape slider defines how closely the AI must preserve shapes to match the original design. Set a lower value if shapes in the input image are not so important.

8 - The Style is a specially trained AI model designed for a specific effect. Such styles are used in AI templates from the Design Gallery to create specific text effects, including balloon alphabet, fur, oil paint, bread, water and many more.

When you use the AI Wizard for the first time, there may be no styles in the list. However, once you open an AI template (free or paid) that includes a style, it will automatically be added to the Style list, and you will be able to use it in future with your own designs.

9 - Advanced settings.

10 - The Size option allows one to select a dimension preset for the AI generated image.

11 - Starts the image generation process. The number on the button indicates how many Render Coins are required to render with the current settings. The generation is possible only if you have enough Render Coins and the connection to the server.

12 - Displays the prompt, seed and other settings used to generate a selected sample.

13 - Zoom in or out.

14 - The preview of the current sample.

15 - Thumbnails of the generated samples. Click on one to preview the corresponding sample.

16 - Export the current sample.

17 - Delete the current sample.

Advanced Settings

1 - The Negative Prompt describes properties that you don't want to have in the generated image. It can be left empty. Example: dull, matte, desaturated.

2 - The Seed is a random number that initializes the image generation process. The Seed can be also set manually. You can copy the Seed from a previously generated image if you want a new image to look like it.

3 - Upscales the current image dimensions. Enabling this option costs an extra render coin.

4 - The Sampler drop-down menu lets you select one of the sampling methods. Samplers that often give good results are: K_EULER, K_EULER_ANCESTRAL, KDPM2Ancestral, KDPM2 and UniPCMultistep.

5 - Sampling Steps defines the number of sampling steps for the denoising process. Higher value increases the image quality but takes longer to render. As a rule, values greater than 30 make almost no further improvement to the output image.

6 - The Universal Control activates the image segmentation. The AI basically tries to define areas in the input image (individual objects like a flower or letter) in order to fill those with proper details. This mode works well in most cases and should be preferred to the Contour Control or Depth Control. In this mode, an input image with less details can still give a good result.

7 - The Contour Control makes the AI find contours in the input image and use those to create the output. This mode works well with "flat" images containing many small details such as a texture. The model tries to preserve those details in the output.

8 - The Depth Control option is often used with input images that have details which are distinctly closer to the viewer than the others. For example, flowers in front and leaves far in the background. With the Depth Control activated, the AI creates a depth map in order to preserve the feel of different distances to the objects.

9 - The Use Refiner option activates the two-stage generation. The base model performs the first part of the task. Then, if enabled, the app runs the refinement model which specializes for the final denoising. The base model does most of the work and the rest is done by the refiner.

10 - Restores the default settings.