Invoke supports the advanced 'compel' prompting syntax, a popular open-source syntax that originated in the Invoke project!
Upweight (+)
Description: The positive modifier (+) can be used in a prompt to increase the focus or attention on a specific term. By adding '+' next to a term, you're signaling the importance of this aspect in the response, encouraging more emphasis on it.
Syntax: term+
Example: "landscape painting, trees++" - This prompt would encourage a focus on trees in the landscape painting.
Note: Alternatively, you can include the specific amount you'd like to upweight a term as a number following parenthesis - such as "landscape painting, (trees)1.2" - Each + multiplies the weight of that term by 1.1
Downweight (-)
Description: The negative modifier (-) can be used in a prompt to decrease the attention or focus on a specific term. By adding '-' next to a term, you're suggesting less emphasis on this aspect in the response.
Syntax: term-
Example: "urban photography, people-" - This prompt would introduce the concept of "people" into the composition, but encourage less significant focus on them.
Note: Alternatively, you can include the specific amount you'd like to downweight a term as a number following parenthesis - such as "urban photography, (people)0.8". Each - multiplies the weight of that term by 0.9.
Blends
Description: The blend feature is used to merge the meaning of two or more prompts together, in order to modify the model's understanding of what you are asking for. Please note - This is not the same as adding prompts together. Instead, this feature is best used to blend in stylistic components.
Syntax: ("prompt 1", "prompt 2").blend(1,1)
Example: ("landscape painting, trees++", "sunset scene, warm tones++").blend(1,1) - This would result in a landscape painting focused on trees and a sunset scene with warm tones, giving equal importance to both prompts.
Was this article helpful?
That’s Great!
Thank you for your feedback
Sorry! We couldn't be helpful
Thank you for your feedback
Feedback sent
We appreciate your effort and will try to fix the article