Well my attempts aren't so promising right now. Can someone give me some hints, what words to use to create a certain result? I trieing to create a Avatar for DnD. A female yuan-ti cleric/priest moon themed a little ^^
I'm not familiar with this character, though I can give some tips for prompting. In the positive write words and phrases that describes what you wish to see in your image. In the negative prompt write words and phrases that describes what you don't wish to see in your image. It is especially powerful to use positive and negative in combination.
Example
Positive: something big
Negative: something small
You have put a very large emphasis on big things.
Adding weight to the prompt.
"(something big:1.2)" you have increased the emphasis by setting a value over 1
"(something big:0.6)" you have decreased the emphasis by setting a value below 1
You can add weight to negative prompt as well.
"(something small:1.2)" you have increased the emphasis on how much you don't want something small
"(something small:0.6)" you have decreased the emphasis on how much you don't want something small
To increase the quality of the image use genre appropriate key phrases that describes quality. If the aim is photo-realism use photography specific terms and camera setting etc in addition to describing the subject. Do the same for stylized images and anime etc.
I'm not sure what would be genre specific terms for anime, perhaps the edge/outline thickness and the strength of color, the amount of texture or lack there of etc.
Also use the style you are going for in the prompt, such as "movie poster" if this is is the goal. In your case: "Avatar image for Dungeons and Dragons". When describing the wanted image don't use too obscure references or terms, nor too technical language. Avoid too subjective terms such as "insanely detailed", instead use "highly detailed" and add weight for more emphasis. It's not about making an "ai" understand. There are no "i" in this. It's "deep learning" and fancy processes and logarithms. So keeping this in mind, always use the more commonly used phrase and it's much more likely to have been used before and therefor existing in the "database".