ᴛʜɪꜱ ʟᴏʀᴀ ɪꜱ ᴘʀᴏᴠɪᴅᴇᴅ ᴀꜱ ɪꜱ ᴀɴᴅ ɪɴᴛᴇɴᴅᴇᴅ ꜰᴏʀ ʟᴀᴡꜰᴜʟ ᴀɴᴅ ʀᴇꜱᴘᴏɴꜱɪʙʟᴇ ᴜꜱᴇ.
ɪᴛ ᴀᴄᴛꜱ ᴀꜱ ᴀ ɢᴇɴᴇʀᴀʟ-ᴘᴜʀᴘᴏꜱᴇ ʀᴇꜰɪɴᴇᴍᴇɴᴛ ʟᴀʏᴇʀ, ɪɴꜰʟᴜᴇɴᴄɪɴɢ ʜᴏᴡ ᴛʜᴇ ʙᴀꜱᴇ ᴍᴏᴅᴇʟ ᴇxᴘʀᴇꜱꜱᴇꜱ ᴘʀᴏᴍᴘᴛꜱ ᴀɴᴅ ᴄᴏɴᴄᴇᴘᴛꜱ ʀᴀᴛʜᴇʀ ᴛʜᴀɴ ᴘᴇʀꜰᴏʀᴍɪɴɢ ᴅɪʀᴇᴄᴛ ᴛʀᴀɴꜱꜰᴏʀᴍᴀᴛɪᴏɴꜱ. ᴛʜᴇ ʀᴇꜱᴜʟᴛɪɴɢ ᴏᴜᴛᴘᴜᴛ ɪꜱ ᴀʟᴡᴀʏꜱ ᴅʀɪᴠᴇɴ ʙʏ ᴛʜᴇ ᴘʀᴏᴍᴘᴛ ᴀɴᴅ ᴛʜᴇ ᴄᴀᴘᴀʙɪʟɪᴛɪᴇꜱ ᴀʟʀᴇᴀᴅʏ ᴘʀᴇꜱᴇɴᴛ ɪɴ ᴛʜᴇ ᴜɴᴅᴇʀʟʏɪɴɢ ʙᴀꜱᴇ ᴍᴏᴅᴇʟ.
ᴛʜᴇ ʟᴏʀᴀ ᴅᴏᴇꜱ ɴᴏᴛ ʀᴇʟʏ ᴏɴ ᴘᴀɪʀᴇᴅ ɪɴᴘᴜᴛꜱ, ʙᴇꜰᴏʀᴇ/ᴀꜰᴛᴇʀ ᴇxᴀᴍᴘʟᴇꜱ, ᴏʀ ᴛʀᴀɴꜱꜰᴏʀᴍᴀᴛɪᴏɴ-ꜱᴘᴇᴄɪꜰɪᴄ ᴛʀᴀɪɴɪɴɢ. ɪᴛ ꜰᴏᴄᴜꜱᴇꜱ ᴏɴ ɪᴍᴘʀᴏᴠɪɴɢ ᴄᴏʜᴇʀᴇɴᴄᴇ, ᴅᴇᴛᴀɪʟ, ᴀɴᴅ ꜱᴛᴀʙɪʟɪᴛʏ ᴀᴄʀᴏꜱꜱ ɢᴇɴᴇʀᴀᴛɪᴏɴꜱ, ɪɴᴄʟᴜᴅɪɴɢ ᴍᴀᴛᴜʀᴇ-ᴛʜᴇᴍᴇᴅ ᴏᴜᴛᴘᴜᴛꜱ ᴀʟʀᴇᴀᴅʏ ꜱᴜᴘᴘᴏʀᴛᴇᴅ ʙʏ ᴛʜᴇ ʙᴀꜱᴇ ᴍᴏᴅᴇʟ.
ɴᴏ ꜱᴘᴇᴄɪᴀʟɪᴢᴇᴅ ᴅᴀᴛᴀꜱᴇᴛꜱ ᴛᴀʀɢᴇᴛɪɴɢ ɪᴅᴇɴᴛɪᴛʏ ᴇxᴘᴏꜱᴜʀᴇ ᴏʀ ᴄᴏɴᴛᴇɴᴛ ʀᴇᴄᴏɴꜱᴛʀᴜᴄᴛɪᴏɴ ᴡᴇʀᴇ ᴜꜱᴇᴅ ᴅᴜʀɪɴɢ ᴛʀᴀɪɴɪɴɢ. ᴀʟʟ ꜰᴀᴄᴇꜱ ᴀɴᴅ ᴘᴏᴛᴇɴᴛɪᴀʟʟʏ ɪᴅᴇɴᴛɪꜰɪᴀʙʟᴇ ꜰᴇᴀᴛᴜʀᴇꜱ ᴡᴇʀᴇ ᴍᴀꜱᴋᴇᴅ ᴏʀ ᴀɴᴏɴʏᴍɪᴢᴇᴅ ᴛʜʀᴏᴜɢʜᴏᴜᴛ ᴛʜᴇ ᴛʀᴀɪɴɪɴɢ ᴘʀᴏᴄᴇꜱꜱ.
ʙʏ ᴅᴏᴡɴʟᴏᴀᴅɪɴɢ ᴏʀ ᴜꜱɪɴɢ ᴛʜɪꜱ ʟᴏʀᴀ, ʏᴏᴜ ᴀᴄᴄᴇᴘᴛ ꜰᴜʟʟ ʀᴇꜱᴘᴏɴꜱɪʙɪʟɪᴛʏ ꜰᴏʀ ɪᴛꜱ ᴜꜱᴇ ᴀɴᴅ ᴀɢʀᴇᴇ ᴛᴏ ᴄᴏᴍᴘʟʏ ᴡɪᴛʜ ᴀʟʟ ᴀᴘᴘʟɪᴄᴀʙʟᴇ ʟᴀᴡꜱ ᴀɴᴅ ᴘʟᴀᴛꜰᴏʀᴍ ʀᴜʟᴇꜱ.
OnlyAnal - Anal Sex - Krea2- K3NK
Lora trained with 3.601 images of women being analy penetrated and vagina closeups
Trained in diffusion pipe at resolutions = [512,720,1024]
Rank 128
Used raw model for training
Increase str up to 2.0 if you dont get anal, is a bit tricky also depends on the bypass lora youre using
Example Prompt
This is a high-resolution photograph featuring a nude, light-skinned Black woman with curly hair, lying on her back outdoors. She is smiling and spreading her legs, exposing her shaved vulva. Her breasts are medium-sized with dark areolas. The background shows greenery and a clear sky. The lighting is natural, highlighting her smooth skin. The camera angle is low, focusing on her genital area. Theres a man's penis penetrating her anus from below.Have fun! 😜
Description
FAQ
Comments (29)
As always, your contributions to the gooning community are greatly appreciated! 🫡
800mb is a bit heavy, but it should give the best results, i might downrank it at some point just for testing
@K3NK I think X penis for Flux 1 Dev was 2 GB or something. I wouldn't worry about 800 MB.
@ss9999 I would worry about it when it causes the results to be bad. Maybe if it was trained less, the results wouldn't be completely deep fried
@pogo i dont get completely deep fried results as you can see in the examples
Maybe she just has an insane amount of blackheads on her leg naturally.
Also, as other have said, WATERMARKS!!! Finding out a top model creator using watermarks in their dataset was the start of my villain arc
@pogo dataset is +3k images, I masked corners and faces, maybe some logos were not fully covered, those "blackheads" are probably the bypass lora overcooking skin.., I've prolly changed my workflow and the bypass lora like 3 times since making that image you linked, but again, are all the examples like that one?
@K3NK looks good bro
finally, a good anal lora that isnt early access bs. thank you my champion
He literally had this on early access for 7k buzz, my man.
Whereas @Loraholic anal helper LoRA does everything you need for less than 5% file size
This is a much more complete LoRA. There is a huge difference between a helper LoRA and a dedicated LoRA.
This LoRA will make it much easier to get the complete package, from scene, to good looking genitals.
Mine is more of a quick fix when you have a image that you like, but the penis goes in the wrong hole.
Didn't a bunch of people pay 7k buzz to get early access to this? And no one submitted any images?
I tested it. Seems alright at lower strengths. Results are noisy and fried at high strengths, as is visible in the sample images.
There is another LoRA that does everything this can for less than 5% the size.
@pogo Which one?
@matai515 Loraholic's Anal Helper. If someone ever mentions a high quality LoRA with low file size, it's Loraholic. GOAT
@pogo Thanks! You're right about Loraholic
Extremely over cooked. Baked in vagina type, face expressions always open mouth or sticking out tongue. Doesn't work well at lower strength either, leads to body horror. Ruins prompt adherence, and to top it off doesn't work half the time without an additional nsfw lora.
You can just look at the sample images to know this LoRA is fried.
Multiple people paid 7k buzz for this when it was early access btw
@K3NK i don't know how you're training but considering your file sizes. Just use ai-toolkit. Use lokr 4, automagic3, lr 1e-4, weight decay 1e-5, weighted or sigmoid timestep, 1024 res only.
Your results will be better quality and probably only need 10-20% of the steps you're doing now.
I get a logo/watermark in the right corner 90% of times & many things go wrong. Better luck next time!
You can see he even submitted a sample image that contained a watermark.
@pushy i masked the four corners of every image (plus faces), since every image had a different logo, and diferent size, probably some of those were bigger than the masking zone (is hard to check all are covered when the dataset is +3k mages), i should make bigger masking zones... i guess i didnt got much of those when testing the lora and i went ahead.
EDIT: seems like using (watermark, logo in the corner:-1) in the prompt + kj prompt weight nodes helps with it. Krea2 seems to be another stubborn model that doesnt listen to negative prompts.. :S
Krea2 NSFW loras too like anal sex without this. At only my prompt? So can make only vaginal sex version? Specially nelson or mating press pose that hard to pussy sex. At my case. :) Thanks to good work.
krea2 is heavily censored in text encoder level, the bypass is a must, you can test this trying to get "anal" written in your image, depending on the bypas lora, usually when used at 0.25 str will refuse to write nsfw texts.. i think anal is a thing very affected by this particular issue..
Doesnt work well for me. Changes the people a lot.
i gues you mean using as edit? you should change the params in the nodes, this lora is trained the same way ive trained all my oter image loras, using automagic, ive mixed datasets pov and other poses, that might be an isue, but changing the subject face is not a lora thing.. xD
Could we get this for Klein9B as well please? thanks for your work.
great!!!!!!!!!!!!!!!!!!