This is not only i2v Lora. The idea is to teach the correct male anatomy.
All examples are t2v.
It should perform way better with i2v if the anatomy is already present, but hopefully when I finish training it will not be needed.
This is a first experimental version. Trained with my experimental implementation of Musubi in my GUI. https://github.com/diodiogod/musubi-tuner_simple_GUI
After proper quality control technique (inspired by Ostris AiToolkit), it maintains quality very well.
I intend to keep training it. It was trained at first with just still images at 256x. Why? Since Klein training first on 256 has been positively good. I'm doing a continuation training on 1024 for details now and maybe I'll try 2048.
I trained 10 epochs, each is 1.629 steps. But I choose epoch 8 as it had the best asshole positioning.
Penis shape is still shit. But it's getting there. If you use a i2v for first frame with a penis it should be great.
No trigger word, just reference a man from behind or showing ass
Some supporting trigger words used on the dataset:
glans, large penis, small penis, medium-size penis, wide open anus gap, closed tight anus gap, penis upside-down, scrotum, testicles, low-hanging testicles, hairy, shaved Description
Epoch 8, so 13.032 steps. 256x. Use 1.0
FAQ
Comments (15)
I've already trained a continuation with 1024 and 2048 res and it's getting way better details. I'll publish as soon as I test it enough and train some more.
Would you consider training a LoRA for male anal sex, h3 really tends to turn the anus into a vagina too easily, and the insertion position is also often misaligned
Not right now as that would require a video training (movement) not just anatomy learning. I don't have the dataset. Hopefully with my next version of male ass the model will at least understand what an anus penis and testicle is and where they should be.
this is half way in next version, but you can do this kind of insertions with just T2V https://civitai.red/posts/30247520
Amazing! :)
Have problems with this like double anus/asshole. Also is misplaced sometimes.
What strength should we use?
Post your example so I can check. Use 1.0. I had double assholes only on the first version, but with proper prompt I didn't see it anymore. In any case, this is still an experimental LoRa.
there's some bizarre shit here that cannot be unseen
then my job on Civitai is done 🤣
I don't wanna live in this timeline anymore.
great!
Have problems with this like double anus/asshole.
can you please post your results with workflow, so I can test. In which version? On current version I didn' get any.
This is glorious. Thank you so much.
And for anyone complaining, if they trained models thoroughly and properly, we wouldn't have to go behind them and train things like this. If you don't like it, insist on models being released in a more well-trained and complete state.
But I love it, well done!
Thank you! I hope than future version will fix the whole male anatomy, not just the ass! We are almost there.