/vt/ - Virtual Youtubers

>>43340613
>>43340732
arigatou.
>>43341036
t. you found it or t. you made it?
>>43341118
it depends on what you're doing. they have their own uses, depending on the input image's characteristics/color - and what you want the output to look like - and even if you're doing txt2img or img2img
TL;DR you kinda do need to RTFM. Not the extension's manual, it's lacking. look at https://github.com/lllyasviel/ControlNet - there are pictures

picrel - HED worked best. Ignore the very first Kaela, I wanted to see what txt2img without Controlnet would make to see if I was actually prompting what I wanted to see in the end result or not. Canny surprisingly worked even though it couldn't see his hands.
Openpose not Openpose hand wasn't able to detect his hand. Might be worth drawing it's "style" of skeleton in bright neon colors on a black canvas in those situations and just feeding it that.