All New Adventures of the Riding Hoods

#16
by o0Rodney0o - opened

Prompt:

Frazetta painted style, a dark elf with yellow eyes and long swords rides through the narrow path upon a large black clidesdale horse, searching, focus on girl hiding behind brush and trees, many dark elves with swords move through the forest, girl in Red Riding Hood clothese, solid black dots (ellipses) for girl's eyes

Preset:

Franco-Belgian

Comic:

Frazetta painted style, a dark elf with yellow eyes and long swords rides through the narrow path upon a large black clidesdale horse, searching, focus on girl hiding behind brush and trees, many dark elves with swords move through the forest, girl in Red Riding Hood clothese, solid black dots (ellipses) for girl's eyes

(I need to be more careful of bad spelling!)

Note: I had entered some of the prompts more as tests to see how they would get interpreted.
Looks to me like they were just ignored.

Hi @o0Rodney0o
Indeed the prompt has a limited capacity

The text you typed, for instance, is 75 token long, which would barely fit into SDXL 77 token length, but in the comic book generator we also need to add:

  • prompt prefix (to ground the story in a specific style or direction)
  • the prompt generated by the LLM

which explains why your instructions were not taken into account, there was just not enough space to fit everything in it

Maybe with future models the context window will be larger

Thanks, that's exactly the kind of information I need to better understand this process.
(I'm new so thankful you are patient with us)

I did realize that shorter prompts tend to produce more targeted output but it's good to see specifics.

Sign up or log in to comment