Support my work on Patreon and Ko-Fi and get access to tutorials and exclusive models.
Rate and leave a like if you enjoyed it, and follow for new models!
How to use:
Weight 1.0 with your usual CFG
Trigger word: minotaur
Other tags that you can use to modify or amplify the effect: orc, monster, nose ring, fur, muscular, size difference, height difference, glowing eyes, fat, fat man, horror \(theme\), manly, etc.
Description
FAQ
Comments (23)
Another awesome lora, ill have to show you some of the results once i get a few bakes in. Do you have an idea for your next lora to make?
thanks! I'm always open to suggestions. and looking forward to seeing your results.
I may have about 2-4 loras going on simultaneously, but in different stages. e.g. I may be gathering data and working on captions for 2 loras, while running trainings on other 2 loras, etc. In the end I post only the few that I'm happy with sharing.
@psoft I would love to see another dildo lora for all sorts and types of dildos, such as bad dragon etc, and maybe having the option for a fuckmachine or something, i have no idea how hard this would be or anything along those lines, im very new to training.
all the Taurusi out there thank you in this birthday month!
lol I didn't even realize that
Man, you're truly a wizard. Can't wait for your next works!
sweeet
more creatures pls xd
ogre, centaur, skeleton
I'm afraid that the answer may be "no", but just in case.
Do you think that there's enough data to make a lora for the gyaru-oh (aka. tanned delinquent)? After your Orc and Ugly bastard, is the last piece we're missing from the unholy trinity of hentai bad guys :P
I was not familiar with the term, but after a quick search, it is basically a tanned blonde male? Any other characteristic? There doesn't seem to be much data out there, but it's a concept very familiar to what the model already knows.
@psoft They are probably the most usual bad guy in NTR hentai nowadays. Common characteristics are: muscular, bleached blonde hair, tanned, piercing/jewelry and gaudy clothes. They tend to have rough facial features compared to the usual main characters (with even some being almost young ugly bastards) although there are exceptions. I picked some sample from you in this image: https://imgur.com/a/jyaO5sE
I guess that the problem is getting enough data, right? How many images would be necessary to train a concept like that?
WONDERFUL LORA! I hope that you will make werewolves someday ^^ would be fun to see your waifu being mated by wolves
I may have something for tomorrow, stay tuned!
@psoft thank you so much! if I may asked, how many pics that you usually use to train many characters like this? I wanna make a new SFW lora that involves many chickens but I am confused about the tagging for multiple characters. I usually make character and costumes lora. I really wanna make chickens lora for SFW projects
@mariegold I always leave the metadata on my loras untouched so that others can learn. You can use the Additional Networks extension to auto1111 or in the extra networks tab.
E.g. click "show/hide extra networks" button next to the Generate button in webui, then locate the lora you want, and on the top right corner of the lora there's a (i) and if you mouse over it you can see the metadata for the training.
In any case, this basically has 262 images, txt captions with "minotaur" as 1st token, and the rest of the captions were automatically generated with kohya gui (but anything that uses danbooru or WD14 tags works). One important thing is to prune (i.e. remove) all tags that relate to the concept minotaur, e.g. horns, muscular, animal ears, fur, furry, etc. This allows the training to relate "minotaur" with the visual elements not tagged.
IIRC, then I trained at LR 0.0001, AdamW8bit, 20 batch, gradient checkpointing, keep_token 1, 512res, 10 repeats, 6 epochs.
But do check the metadata on the lora for better precision on this info
@psoft thank you! now I just need to make 262 chickens, I usually use Blip captioning for caption. Maybe I can try using danbooru tags or WD14. Had tried danbooru before and I can't wait to many chickens :3
@mariegold sounds great! :D you probably can make it with around 100-150 images
@psoft Thank you sensei. Now we can both make Colonel Sanders proud
Really interesting LoRA. I found that you can actually create a female character with minotaur horns. 5/5 as usual
thanks! you're right, depending on the training and base models, LoRAs can be used very creatively outside the initially intended outcome :)
Cant get it to work and im not sure why
even replicating all the generation data of the example images?
@psoft USER ERROR, I apologize im an idiot.
Hello, i found this lora on a pony version, but i can't find the source from the image, so i guess it's probably from this lore, do you know if is it possible to turn this Lora into a Pony version?
Edit: yes, it's possible, just need to click on the SDXL option on the Lora (From the wrench and hammer icon at top right) and worked.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.