How do picture prompts work?

Image to image can work together with text prompts and picture prompts to generate the image you want. There are three referencing modes for image prompts:

1. General Reference:Images are general prompts and work with prompts (optional).
2. Face Swap:The image is a portrait of a person, and the generated image will reference the face information of the person in the image.
3. Structure Reference:Reference to the posture of a person, or the outline of an object.

Examples

1. General Reference

Images are used as hints to generate images in different styles. In the image below, an anime image (style "SAI-Anime") is generated from the girl's image, in which cherry blossoms, girls, and the girl's dress and posture are all referenced.

In the following figure, we provide 2 pictures, a photo of a lake and a photo of a castle, and prompt "Mountain, Lake, Castle" (the style is "Terragen"), then the prompt and the picture work together to generate a picture that is very similar to the picture provided.

2. Face Swap

Provide a portrait of a person and prompt "A girl sitting on a couch, reading a book, a fireplace" (select the type "Portrait"), the resulting picture will reference the face in the provided picture.

3. Structure Reference

If you provide a picture of a mountain peak with the prompt word "China's mountains, the Great Wall, and the beacon tower" (the style is selected as "Terragen"), the outline of the main peak in the generated picture and the mountain in the prompt image are very similar.

Example: How to set the type, style, and prompt word