Model Guides

Introduction To Model Guides

After learning about the basics of prompting in our Prompting 101, you can learn more about your favorite models using the guides below! We'll continue to add additional guides for other models, so keep checking back for more!

Guide Sections Explained

SectionExplanation
What This Model Excels AtDescribe the model's strengths and unique capabilities.
What To Be Aware of This ModelHighlight any limitations, biases, or potential issues
Prompt StyleExplains the recommended way to structure prompts for this.
Sample PromptsProvide ready-to-use examples to interact with the model.
Prompt Style ComparisonHighlights the differences between the suggested prompt style and the general prompt style.
Final ThoughtsA summary of key takeaways and addition tips

Prompt Testing Process

For consistency, every model is first given a test prompt using the [Automatic1111] prompting style for both positive and negative prompts, referred to as the generic or universal prompting style. Every image includes at least the following prompts.

Positive

Closeup portrait shot, a woman standing in a forest, masterpiece, best quality, extremely detailed, soft light, soft shadows, soft backlighting, (best image,best quality:1.5)

Prompt ComponentsDescription
closeup portrait shotShot statements that sets the image to be a  side face portrait of the subject.
a woman in a forestsubject statement.
masterpiece, best quality, extremely detailedQuality statements that guide direct the engine into creating a well composed, high-quality,, detailed final image
soft light, soft shadows, soft backlightingCombination of lighting terms that create a soft glowing light, fuzzy shadows and adds backlighting to give it dimension and depth.
(best image,best quality:1.5)Best keywords help select the best guess of the image during rendering

Negative

(low quality, low detail:1.5),(worst image, worst quality:1.5)

Prompt ComponentDescription
(low quality, low detail:1.5)Keeps anything image element that to far away or blurry out of the final image.
(worst image, worst quality:1.5)Set the Model to read the final image and ignore the worst possible choices
Testing Steps
Then the original prompt is modified based on engine specific behavior to improve the result, usually by changing the word order or slightly modifying the supplied prompts.
Finally, the guides present a comparison of the the general and recommended prompts to discuss the differences between them and how using the recommended prompting style will influence your results.
This assessment is based on multiple tests of the model summarized for purposes of the guide and considers the inherent strengths and weaknesses of the specific model.