[Please comment to report Problems or for Suggestions to improve this Lora, or which Checkpoint you want it trained on]
An expermiental try to asure the "folded back" position of animal ears of anthro wolfs/foxes/cats.
Trained on anthros made with "pony realism", but trained on the standart Pony model.
Used with the Tags: "Frontview", and "sideview", to better controll the ear perspective.
like "folded ears, sideview"
or "dropped ears, frontview"
suggestet strenght: 0.8-1.3
"Ears folded back", "Folded ears", and "Dropped ears" seems the strongest now, the later makes the ears sligghtly more hanging sideways. "
EFB-1.0 Pony, works great with "Pony" and "Styles for Pony" especialy watercolour.
Beta 2 has much less noise and is usable without High-Res-Fix, suggestet strenght: 0.8-1.3
Beta 1 is trained with to much tags and very noisy. suggestet strenght: 0.7-0.8
Description
Version best to use with classic pony+styles
FAQ
Comments (6)
results inconsistent, hard to tell if the lora is doing anything at all. Get similar results without lora at all, similar success rate
Thank you for your feedback!
I know this Lora has a lot of room for improvement, and i appreciate your comment.
I tried a little bit with your Prompts and your Checkpoint.
"Ears folded back", or a higher strenght makes the ears alittle bit more rotated.
Dropped ears are mostly sideways unlike the very flat ears of an angry cat.
The very strong folded ears like of an angry cat, is yet very difficult and rare with this lora. In the training data there are some photos of cats with this ear postion, and under the prompt "realistic" they sometimes emerge.
But this Lora seems weak for Fantasm-Pony. I tested some of my work on this model and it has some effect but not as strong as under the original pony. The Compability with other Pony-Checkpoints is an big Issue my Lora has.
Maybe an other contributing fact could be, that this Lora is not trained on anime characters or more humanlike characters like nekos, but this is only a suggestion.
I collect nowadays data for a new run, and i know i have a lot to improve on prompting, checkpointcompability and strenght of expression.
Thank you for your help and posting on this lora!
@BrettBoard we'll get there eventually, sdxl still has issues to iron out but there's some room to grow yet
IL version pls
ok i start to work on soon^^
done!
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.