A Femboy 'Concept' LoRA. Turns the subject(s) of a generated image into an erect femboy/crossdressing femboy virtually without fail, and creates a strong lean towards:
Visible assholes/gaping assholes 'before' and 'after' sex.
Lean-yet-toned body/androgynous body aesthetic.
Anal sex with an off-screen partner.
Group sex with the femboy subject(s) as the focus.
Having the subject(s) wear heels, toeless legwear, chokers, and ball bras.
Demure posturing such as a hand on one's chest, and presenting with spread legs.
It cannot do flaccid penises or chastity/it has very little training with them.
Important Use Notes:
You must use the activation word 'femboi' in your prompt, and ideally it'll be the first word (its strong enough to replace girl/1girl/femboy entirely for labeling a human subject).
Use EasyNegative in your negatives (Repository here).
Highly recommend that you do gens in dimensions between 512x512~640x640, and then use hires fix to upscale+re-gen them, as it has far more success making coherent images within these ranges compared to 768 and beyond.
This lora should function well at 1.0 strength, especially in simpler prompts, but you can try adjusting it 0.6~1.2 for effect, as well as tweaking the keyword directly (such as femboi:1.1).
Adding 'bent over' or 'from behind' as 2nd in the prompt with 'femboi' as 1st should consistently generate ass shots/'cocktails'.
Use 'open-toe high heels' or 'wedge sandals/sandals' rather than high heels, as the lora has been trained to work with open-toe footwear specifically.
Adding 'spasm' near the end of a prompt will cause those 'currently orgasming' lines to appear around the subject(s).
Adding 'heart eyes' will cause accurate heart icons to form around the subject(s), and when used along with 'spasm' they will further increase the chance and accuracy of an anal sex image. 'heart' as a word alone in the prompt will just add the accurate heart icons.
Adding 'logo:2' as a negative to a prompt can augment images. I say augment because it isnt guaranteed to enhance quality and sometimes just changes the image, but it can and may be more effective on images that seem weaker than average with art quality or composition.
This lora was trained on CounterfeitV2.5, but can likely offer solid performance on other anime-style models (my comment post near the start of comments on this lora has a quick comparison of various models).
Recommended Parameters:
Clip Skip: 2
Sampler: DPM++ SDE Karras (other samplers should still work though)
Step: 50 (20~50 can still work, as well as higher)
Scale: 9 or 5 (3.5~12 is a good range)
Text Prompt (the tag order affects it): femboi, masterpiece, best quality, beautiful, fine, delicate, -You should add more details starting here if needed- <lora:femboiFull:1>
Negative#1 (in this exact order): worst quality:2, low quality:2, EasyNegative:1.14
Negative#2 (in this exact order): worst quality, low quality, EasyNegative:1.14
Negative#3 (in this exact order): ((EasyNegative)), worst quality:2, low quality:2
Hires Fix Parameters:
Upscaler: R-ESRGAN General WDN 4xV3 or R-ESRGAN 4x+ (you may need to enable them in 'Settings > Upscaling')
Upscale by: x2 if using 512x512~640x640, according to what you can handle.
Hires steps: 25~40 (the more you do the cleaner the image will be, but also gradually loses some vibrancy)
Denoising Strength: 0.42 (should keep genitals/hands from spazzing out most of the time. Consider lower to preserve it more, and consider moving the original gen-step amount up or down one to shift things slightly if needed)
Originally this started as a very small and personal attempt at creating a lora to help gen femboys better, that a close friend had encouraged and guided me through doing. Since it ended up becoming something I find to be very useful, I'm sharing it here in hopes that it helps others create amazing femboy art, and/or serves to show what to pursue and not pursue when creating your own femboy lora or model.
Description
Important Use Notes:
You must use the activation word 'femboi' in your prompt, and ideally it'll be the first word (its strong enough to replace girl/1girl/femboy entirely for labeling a human subject).
Use EasyNegative in your negatives (Repository here).
Highly recommend that you do gens in dimensions between 512x512~640x640, and then use hires fix to upscale+re-gen them, as it has far more success making coherent images within these ranges compared to 768 and beyond.
This lora should function well at 1.0 strength, especially in simpler prompts, but you can try adjusting it 0.6~1.2 for effect, as well as tweaking the keyword directly (such as femboi:1.1).
Adding 'bent over' or 'from behind' as 2nd in the prompt with 'femboi' as 1st should consistently generate ass shots/'cocktails'.
Use 'open-toe high heels' or 'wedge sandals/sandals' rather than high heels, as the lora has been trained to work with open-toe footwear specifically.
Adding 'spasm' near the end of a prompt will cause those 'currently orgasming' lines to appear around the subject(s).
Adding 'heart eyes' will cause accurate heart icons to form around the subject(s), and when used along with 'spasm' they will further increase the chance and accuracy of an anal sex image. 'heart' as a word alone in the prompt will just add the accurate heart icons.
Adding 'logo:2' as a negative to a prompt can augment images. I say augment because it isnt guaranteed to enhance quality and sometimes just changes the image, but it can and may be more effective on images that seem weaker than average with art quality or composition.
This lora was trained on CounterfeitV2.5, but can likely offer solid performance on other anime-style models (my comment post near the start of comments on this lora has a quick comparison of various models).
Recommended Parameters:Clip Skip: 2
Sampler: DPM++ SDE Karras (other samplers should still work though)
Step: 50 (20~50 can still work, as well as higher)
Scale: 9 or 5 (3.5~12 is a good range)
Text Prompt (the tag order affects it): femboi, masterpiece, best quality, beautiful, fine, delicate, -You should add more details starting here if needed- <lora:femboiFull:1>
Negative#1 (in this exact order): worst quality:2, low quality:2, EasyNegative:1.14
Negative#2 (in this exact order): worst quality, low quality, EasyNegative:1.14
Negative#3 (in this exact order): ((EasyNegative)), worst quality:2, low quality:2
Hires Fix Parameters:
Upscaler: R-ESRGAN General WDN 4xV3 or R-ESRGAN 4x+ (you may need to enable them in 'Settings > Upscaling')
Upscale by: x2 if using 512x512~640x640, according to what you can handle.
Hires steps: 25~40 (the more you do the cleaner the image will be, but also gradually loses some vibrancy)
Denoising Strength: 0.42 (should keep genitals/hands from spazzing out most of the time. Consider lower to preserve it more, and consider moving the original gen-step amount up or down one to shift things slightly if needed)
FAQ
Comments (27)
fantastic!
what 200 epoch 96000 steps ? how many images and repeats did you use o.0
thats 480 images x 1 time ?
@barderino Yeah~ In the the beginning I had alot less images and ran it at 1000~1500, but quickly I started to understand that if I wanted it to make the kinds of images I would be satisfied with consistently, I was going to need to train the lora to produce not only penises on the 'subject' but certain poses, hairstyles, male penises/male chars meant to sex the subject, and other misc things. It had alot to learn, which is why I started trying much higher training amounts.
@AzaTech mh have myself a huge dataset tho don't know if is worth to train a lora with so many images
@barderino I'm not anywhere near an expert on these things, but I imagine it depends on how many things you want incorporated into the lora and how varied they are. Something like a long purple braid and blue eyes can benefit from more angles but seems simpler to portray correctly compared to something like forming penis shapes correctly + placing them around the subject correctly + having cum from them and fly towards/land on the subject correctly + go into the subject's anus correctly + placing one on the subject correctly that is a smaller size/shape compared to the others, and being able to do all this even if there are multiple subjects.
The more I worked on this the more I felt that having the lora be able to portray correctly rather consistently and not just a 1/80+ occurrence was going to take alot of training along with a large dataset.
Bless you, AzaTech! You have brought a smile to my face and blood rushing to my midsection. Thank you, thank you, thank you!! 😘❤️❤️
I'll be sure to try this out soon, and I will review when I get some experience with it.
Nice pics very realistic and exciting
works well with photo real models as well
Totally unusable for me for some reason. Every result is a mess. The other "(Concept) - Trap/Femboy/Otoko No Ko" lora worked fine, but now deleted from the site. Looks like I'm left with only "Esthetic Futanari + Trap" option.
Since it was trained on counterfeitV2.5, try using that model to start. Make sure to actually add the word 'femboi' to your text prompt as well as the lora name, and try following some of the example settings mentioned in the description on this page.
@AzaTech I did add the recommended prompts according to the instructions in several variations, to unsatisfactory results. I'll be honest, I didn't try counterfeit 2.5, but I'm not a huge fan of being locked into one usable model. I've seen people in the comments using different models and it's NOT NEARLY as bad as my results. Mine are straight-up broken.
On the sidenote, NSFW is not really my aim, I just want SD to undertand what femboy is.
Update
Tried to explicitly copy all the prompts and setting. Found out that trying to excede resolution of 640 produces really bad results every time. Also sampler matters quite a bit, DPM++ SDE Karras producing waaaay better results. Toning down the lora to about 0.5 also helps. I'll try to get something decent and post.
@Miraihi If you click on the 'show more' in the lora description virtually all of this is explained under a header called 'IMPORTANT USE NOTES', yes. You definitely want to gen at 512~640x640 and use the hires option to re-gen the image at a higher resolution rather than trying to gen directly at higher res.
@AzaTech It's still extremely inconsistent and produces broken results sometimes, and screws color palette. I've never seen Lora with such strict use cases. I have no idea what's exactly the cause, but looks like I have to put it aside.
pretty much the same problem here.
completely screws the color palette..but I clearly remember having used this lora a month back and it was pretty much perfect.
Now it screws the color palette and gives me sometimes extreme chromatic aberations :/
Same thing for me, total nightmare fuel results.
@Miraihi yeah im having same issues, it is so extremely strict and even then the results aren't great, or not nearly as good as i see in other examples
i like this ero lora,but there is usually a penis in ass,I want to beautiful anus
Thanks for sharing , and thanks for the video , havent heard that song in years , made me nostalgic haha , really good lora !
When using the "from behind" prompt, every image is getting a massively gaping anus. No matter what I do, I can't seem to get rid of it. Any suggestions?
Hmmm, you can try turning down the lora strength very slightly in increments, as in changing the <femboifull:1.0> to <femboifull:0.98> and 0.96 and so on - you can also try doing this to the femboi tag itself or the from behind tag, such as femboi:0.9 and see if that helps - it may be a battle of rng to a degree as well, because if you go too low youll start to generate penises less in the from behind view
@AzaTech I'll give it a shot, thanks!
A lora doesn't need 96,000 steps. It can accomplish what you're trying to do at 1,200-3,000 steps. It won't listen to me when I tell it I want shorter hair, it doesn't listen to solo prompts, it is hell bent on making the femboy fuck... it isn't that I don't want the boy to fuck i just need it to be when I need him to.
Im having a really hard time getting these to work. idk why. the from behind does not work, and the images come out low quality... idk what im doing wrong
Wow how did i never see this lora! This is great to get dick-having humans. It's so hard to get rid of vaginas...
Easy, host a Smash Bros tournament.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.



















