Then let me give you another hint. SDXL and so PonyDiffv6 use clip skip 2 by default. If you really want a certain aspect to be present in a pic describe it multiple times with slightly different wording. InvokeAI even is going so far as to have a feature to multiply your prompts for that reason. In comfyui you have to do that by hand.
I noticed that because I use “realistic horse legs” as a prompt when I don’t want the tube like legs of the show. Sometimes the pic gets more “realistic” sometimes it gets more “horse” and sometimes even both. And sometimes I get the result I want, more often then not actually.
@Background Pony #84A4
Probably did that because of the friendly tag I used. Pony Diffusion v6 is weirdly hung up on some things. If you use a verb to describe a specific aspect of your picture you can change the whole outcome with that. You need to be careful with that. Like, drop “horse” anywhere and you bet all your result look nearly like horses with a Twilight wig.
“Thick muscled anthro Spike” Buddy boy, the last episode of FiM made him into basically a scaly gorilla, i think that description is putting it generously.
Also as for the art its self, the AI did a surprisingly excellent job with the only issue i can see being the eyes look off, like they don’t fit the style somehow.
That’s exactly the result I want, thanks for the hint.
Probably did that because of the friendly tag I used. Pony Diffusion v6 is weirdly hung up on some things. If you use a verb to describe a specific aspect of your picture you can change the whole outcome with that. You need to be careful with that. Like, drop “horse” anywhere and you bet all your result look nearly like horses with a Twilight wig.