>>96059551
Every ones mileage may vary, but this is a good starting point.
>The general shape of a prompt matters a lot. Unlike the text to search or search to text models, image gen wants absolutely no sentence fluff. You aren't writing sentences, you're writing states and objects.
>The general structure of a prompt should be the following: Style, location, object, state[, object, state][,object, state] etc. Each time you add a object though the harder it gets on the AI. You also want to cut basically every preposition or conjunction out of the prompt except a couple words that pull a lot of weight in English like the words, And and By. By can be replaced by other distance qualifiers like nearby, beside, next to, away from, but you really have to test to find out what works.
>Nouns and adjectives are the only things in the object portion and make up the bulk of the prompt. State is the only place to put verbs and adverbs and needs to be short and simple. Use proper adjective construction OSASCOMP for best results.

https://files.catbox.moe/xpk9c9.jpg