This wildcard adds a list of basic adverbs (words that modify a verb) to provide additional variations for prompts. These are intended to be used with prompts describing emotions or facial expressions, but they should work on any verb. All of these adverbs end in "ly," so expect "passionately" or "angily" results as you write your prompts. I'm not sure how much of an impact adverbs have on image generation, but it does seem to change the results in a meaningful way.
Please enjoy!
Description
FAQ
Comments (2)
Quick question on the prompt structure.
How should I structure the prompt, do I add "__Emotional-Adverbs__" to the prompt text? e.g., "__Emotional-Adverbs__ couple furiously having an argument",
or is it "couple furiously having an argument __Emotional-Adverbs__"
or would it be "__Emotional-Adverbs__, couple furiously having an argument"?
While I'm inclined to suggest the former option, I suspect that it doesn't actually matter very much. It's unclear just how sensitive SD is to adverbs, or if they were even a part of the original training data. It may be these wildcards are only effective because they change the spacing and focus of the tokens. My suggestion is to try both, and if you notice that one has a more observable impact, PLEASE report back with your results.