Difficulty prompting specific yoga poses

Hello, Could someone with prompting experience help with advice on how to image gen. specific yoga poses? It seems to be very difficult to do that or I don't know any tricks. For example, picturing an instructor doing a wide leg forward yoga bend but the perspective is from behind. But because he's bending forward completely, his face is between his legs and facing the viewer.

It seems this is too complicated for perchance or most image gens to generate.

3 points · 3 comments · view on lemmy.world

3 Comments

Grimy@lemmy.world · 1 pts · 204d (1 reply)

I don't know perchance but what you need is something that can use controlnet with poses.

https://docs.comfy.org/tutorials/controlnet/pose-controlnet-2-pass

Uncouple3194@feddit.nl · 1 pts · 204d

thank you that looks complicated. I think I was asking about perchance users specifically. And if any of them maybe has some tricks to get it to generate certain yoga poses, like the one I suggested above.

yotv@lemmy.world · 1 pts · 201d

It really sucks at that stuff you can try an image to prompt tool like this https://imagetoprompt.org/ since some stuff does work on perchance but it can be hard to describe it in a way the ai understands. but many uncommon positions will turn into a glitchfest regardless or the image to prompt can't tell what is going on. and if you are desperate you can use the reference image on another site with image to image like this https://www.nsfwlover.com/nsfw-image-edit and then edit the image until you get what you want but this sometimes fails if you make too many changes