REAL DESCRIPTION PART:
Previous version were made based on search with Size Difference tag Huge Insertion/Impossible Fit search basically end up with very similar concept but completely different data set and tags.
Huge Insertion: This time search based on huge insertion + stomach bulge tags rather than size difference, so it has completely separate data set than previous versions.
WAN 2.1 Dragon is T2V model trained on Civit was the experiment that seems failed, for now i can't continue WAN loras since is not possible on Civit and i don't have local set up that is even near close to training local loras.
I saw there are Bigger dom and Human Onahole models, but wanted to try something that will make dom even bigger.
Most of images i found for it contained 3 groups of creatues:
Humanoid (i put here evrything that had hands)
Dragon (obvious, i really wanted to tagg it "bad dragon" but i'm 99% sure that it is alredy functional tag on site)
Beast/Monster - evrything that was not humanoid or dragon
Changelog: V3.0 for IL
has much bigger data, V2 got 20 per monster/dragon/humanoid
this one has 60 of each,
trained on WAINSFW 12 instead of base 0.1 IL
Added onahole "pose"
Changelog: V1.0 was breaking faces way too often so i tried once again.
Added tags: master piece / best quality mostly for stuff from studio fow / D-rex.
added tags: bad quality, bad face, low quality face, badsubface, for negative tagging.
removed some pictures that i considered bad quality.
Trained without beast this time to see how it changes things, just humanoid and dragons.
removed the typos
The best way to support me is to subscribe or write some comment on my
YouTube - Where i shitpost almost everyday
My socials:
DISCORD SERVER
If You want to contact me Discord/Civit are best options since i check those everyday!
Twitter FB
My loras are available from my official accounts on:
Tensor Art: https://tensor.art/u/836328663648595213
CivitAI: https://civarchive.com/user/Shrekman17
Seeart: https://www.seaart.ai/user/ab14b528c289f390e228b8c5d6f497f1?u_code=DAS7ETB