I present to you my second LoRA for a concept/pose that I could not find. This pose was simply not possible (at least for me) to generate with pony models or otherwise by prompting for it.
EDIT: Enabled early access at a very cheap cost to fund the upgradation for all of my LoRAs to Illustrious/NoobAI.
Version 2 Usage Info (Illustrious):
USE AT 0.8 STRENGTH.
Keep the trigger words towards the beginning of the prompt (not always required, but recommended). Reduced the required number of keywords with v2, and should work even better than the Pony versions. I knew very little about training LoRAs when I made v1. Now it should be way better. Enjoy!
Make sure to use these in your prompt (ILLUSTRIOUS/NOOB only. Pony triggers are different - refer to v1):
POV_spooning, on_side,Add "facing away" if you want the character to not look back in this pose.
The showcase images should also give you a good idea of the recommended prompting style.
Version 1 Usage Info (PONY):
This LoRA should work best with 1344 x 768 resolution. But, this time you can use 1216 x 832 (which makes it possible to use with the onsite generator, with lower certainty of the pose).
I use it generally at a strength of 0.9 along with other character LoRAs. But for best results you might have a play a bit with the ratio of LoRA weights between the character LoRA and this one. Start from 0.85, go upto 1.0. It does get wonky sometimes with the bending angles. Go down to 0.65 for more "creative" results. (This doesn't apply to v2)
Trigger words for v1 (Pony):
POV sp00ning, back view, on bed, on side, penetration from back, POV 1man, POV from back, lying If you want to support more models like this, or you just like what you see: ko-fi (｡♥‿♥｡)
Recommendations/Observations (Does not apply to v2):
If you start to get weirdly bent anatomy, try removing the facial or frontal feature trigger words of your character LoRA in your prompt. If anything else doesn't work, please let me know.
When using realistic pony models, turn down the strength to 0.7 or even lower. Seems to work better that way.
If you are using A1111/Forge, and the LoRA does not show up, you can force it to appear by enabling the "Always show all networks on the Lora page" option on the settings page. (Thanks to @pr0log for pointing this out).
Description
Add "ass exposed" trigger word if it gets covered.
Add "vaginal penetration, asshole" if you want vaginal sex instead of anal, which the LoRA tends to lean towards.
FAQ
Comments (9)
Neat concept. Tried to do it locally with Forge using Autismmix and Forge doesn't recognize it. I tried a couple of different models from pony to see if it was just autismmix and it didn't show up on either. Is it just me or am I doing something wrong?
Doesn't recognize it as in it doesn't show up? I'm sure you already put it in your lora folder?
Btw I haven't used forge ever, only Comfy. So could be forge thing. I would wait for another forge user to check whether this works or not. Don't see any reason why it shouldn't.
@nottheninja Yeah is in the lora folder. even rename it. It might be a Forge thing, happen to some other Loras from Civ, but not as frequent. Thanks.
This is correct, Forge (and likely base A1111) doesn't recognize it as a Pony Lora. You can still force it to appear by enabling the "Always show all networks on the Lora page" option on the settings page. I assume it's some setting or option when training. I didn't test it much, but the Lora does still seem to function as expected.
@pr0log Thanks! that did the trick, setting/extra network and I clicked on the "always show all networks" and now it appears and works!
I'm pretty sure the real problem is that it's labeled as SD1 in the metadata's SD Version.
I loaded a 1.5 model, it appeared in the LoRA list, opened up the metadata and set it to SDXL.
@SomeSloppyBoi Didn't know you could do that. Thanks for the info, appreciate it!
anyone knows which checkpoint/model is this creator using?
I used T-ponynai3 v6 for most of the showcase images. Always feel free to comment on any image of mine to know all the gen details. The images should also have comfyUI workflows attached to them, if that helps.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.