Stable Diffusion Models offer a number of styles for the image output.
 |  |
Style is a crucial part of the prompt. The AI model, when failing to recognize a requested style, usually defaults to one most common in related images.
For example, given the core prompt of “landscape,” the model would likely generate landscapes that were realistic or in the style of an oil painting.
Having a well-chosen style together with an effective core prompt is sometimes enough to create a fully-realized concept; after the core prompt, the choice of style influences your final image the most in a simple prompt.
The most commonly used styles include:
- Realistic
- Oil painting
- Pencil drawing
- Concept art
There are a number of ways to invoke a style in your prompts.
To take an example from above, the following are ways you might format a prompt for a realistic image:
- a photo of [core prompt]
- a photograph of [core prompt]
- [core prompt], hyperrealistic
- [core prompt], realistic
You can, of course, combine these modifiers to pursue greater realism, but a little often goes a long way.
Here’s our panda from Section 1 in each of the prompts above:
Referring back to our style examples, for an image in the style of an oil painting, adding something like “an oil painting of [core prompt]” to your prompt works well..
This sometimes results in the image showing an oil painting in a frame, to fix this you can just re-run the prompt or use negative weighting (discussed below).
“An oil painting of a panda”
For an image in the style of a pencil drawing, one easy approach is to add “a pencil drawing of” to your core prompt or make your prompt “[core prompt] pencil drawing”.
“a pencil drawing of a panda”
The same applies to landscape art.
“A landscape painting of a panda habitat”
