Version two of my second LoRA training attempt modeled off of the Puazi from Incase Art's fantastic Xeno Biology and Puazi CYOA series. This LoRA version is trained using the Anything v4.5 checkpoint, and is compatible with all models derived from that base. I previously didn't advise running this with SD1.5 models, but it actually runs very well on more realistic models.
A few considerations when using this LoRA:
This LoRA, like many others, is incredibly powerful, and can lead to over-weighted results really easily. I do not advise using it above 1.0 weight. With the lower end of the 0.3-0.6 range, you will get results that are mostly in the style of Incase's Puazi, but will look more like orange goblins, while on the higher end you will see more of the unique physical traits of the Puazi come through. Play around with it to determine what's best for you, but 0.7 seems to be a good starting place.
This model has a bad habit of generating two, or multiple puazi in a scene, so add things like "multiple bodies, multiple puazi in background, puazi woman in the background" in the negative prompt to discourage this (will not be 100 percent effective, but was the attempted improvement over V1).
Consider trying GtsuyaStudio's Incase Style Lora at 0.5-0.7 ranges alongside this model for interesting results.
Use the Anything 4.0 VAE as well for slight improvements of color and look.
You still need to be specific to accurately recreate the look of Puazi, so adding things like "heterochromia, blue and orange eyes, skin markings, pointy ears, horns, scarlet/orange skin, blue vagina" etc. to the prompt will help the model accurately match what the result should be.
This model really loves added appendages at higher weights. I will try to fix this in the future, but it is what it is currently. Keep generating to get lucky.
I have not attempted to create femboy/futa characters with this model, and the source material I used does not feature any of the characters with these features.
The model seems to always generate blushing Puazi. Not entirely sure why this is, but just a note.
If you enjoy this model, I would love to see the results you come up with. I personally have long enjoyed the Puazi comics and have hoped for more for a long time.
I acknowledge I do not own the original work of Incase, nor the intellectual property of the Puazi characters.
Description
FAQ
Comments (4)
did you get permission for this? i could see it as a good virality method for incase to spread it's brand more..
I would love to see this used as another tool for them to create. The best AI image still lacks something. Now imagine if Incase could use these images as a base and then "finish them". Like other programs ,Gimp, Blender etc. Instead of taking X time to make a comic, they could make them in a fraction of that time. Heck imagine if incase made images with Ai in mind, and could tailor the learning to be exact what they want, because they made the exact image for that prompting. Ai keeps twisting the body? Here, this is EXACTLY How it should look.....
Rant.* I really wish we viewed things in terms of "tools" to make work easier, and less as way to replace a person.
A surprise to see somone make this, but a welcome one!
FUCKING LOVE PAUZI. Now if i can get the Bible black and this to play nice. I will post, and we can all bask in its greatness. Thank you for this. Pretty sure you just got brownie points from the bouncer at the pearly gates.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.













