Quoted By:
I wish there were a way to signal to the AI "do not subword tokenize this" in a prompt.
Like when you want "screen_light" or "reverse_bunnysuit", it's going to send the individual words (and even subwords like "bunny") which won't get you the right tag.
I tried implementing it myself in modules/sd_hijack.py and modules/promp_parser.py but "is_split_into_words" didn't work the way I guessed it would.