>>44794024>depth tells the ai what should be where, with black meaning 'do whatever', shades of grey meaning 'prefer to put something here', and white meaning 'very important!!!!'...no, depth is meant for use with depth maps. that's a side effect
>canny is similar to depth but needs more tard wrangling. more accurate to source material thoughcanny and HED are shit like depth. they are edge detection algorithms.
depth and normal can "see" in 3D. canny and HED strictly "see" 2D
>none of the ohter models matterMLSD has its uses. Not here. It's more useful than openpose to me.
>>44793762Want to link an unanswered one?
>>44794364None whatsoever.
>>44795231kys
>>44796180>>44796268that orange box is resizing modes, if you don't pre-crop your image the way you want it / resize it to match the aspect ratio of your gen
>Envelope (Outer Fit)Will add a border, the AI will fill in the blanks there while the image generates
>Scale to Fit (Inner Fit)Will crop your image to achieve the correct aspect ratio
>Just ResizeIf you're a fucking psychopath that stretches 4:3 content to fill a 16:9 display and sees nothing wrong with it, this option is for you.
secret: if you "preview annotator result", you can
1. clean up the lines for canny/hed manually to "help" the AI out
2. or just save that image, remove your source image from controlnet, upload that instead, and then turn off the pre-processor and save yourself a bit of time if you think you'll be re-running gens 