Introduction:
Erect Horse Penis LoRA for all humanoid characters. Trained for best compatibility with anime models only. Would work alongside with most character, concept or style LoRAs, except ones, that already have trained penis concept.
Please, check the description, there is plenty of useful information.
Flaccid Horse Penis LoRA - Flaccid version of horse penis LoRA
Recommendations:
1.0 emphasis
See relevant tag tips
Relevant tag tips:
There is a nice feature in prompting for Stable Diffusion - emphasis. Using this, Large Penis and (Large Penis:1.2) will do different sized penises. Second one can be just bigger than large, or even similar to huge penis. Don't use emphasis value higher than 1.4.
You can combine tags below as you like to get the result you need.
Smallest possible penis / testicles - Small, the stallionis maximus - Gigantic
Erection|Size Penisfor penis state and sizeSizevalue can beSmall|Large|HugeorGiganticI do recommend to start from
largesizeYou can use two tags at once to achieve bigger result
"small" tag is highly unstable due to very small dataset, ironic
Size Testiclesfor testicles sizeSizevalue can beSmall|Large|HugeorGiganticI do recommend to start from
largesizeYou can use two tags at once to achieve bigger result
"small" tag is highly unstable due to very small dataset
Sagging Testicles|Resting Testiclesfor testicles stateSaggingfor saggy testiclesRestingfor on-surface state (like breast rest)
Cock Sheathfor proper penis sheathFlared Tipfor flared penis tip shapeYou can use it in negative like that
(flared tip:1.2)to make penis more blunt
Thick Penisfor penis girthMay be unstable, need further improvements
Precum, cum, wet, suggestive fluidin your negative prompt for less mess.Navelfor penis be less in "pointing up" position, while naked
Autopaizurifor penis between breasts (pre-v5)Don't use this tag alongside with
ErectionNot implemented in
version 5or higher
Autofellatio+Oral | Blowjob | Licking Penisfor self fellatio (pre-v5)You may use all three fellatio tags for higher change of autofellatio
Don't use this tag alongside with
ErectionNot implemented in
version 5or higher
Skin-colored Penis|Testiclesfor regular skin-toned endowmentDark Penis|Testiclesfor more dark penis or testiclesColor Penis|Testiclesfor your color
Skin-colored/Dark/Colortesticles also changes sheath colorSheath color can be changed independently, highly unstable
Some colors may perform better than others. If you're getting the same color every time, just add it to negative prompt. Never forget about negatives. It would do the trick.
Google Docs page - older prompt tips (v1 - v3)
Example penis prompts:
Use this if you have no idea how to use this LoRA and didn't read prompting tips above.
For small erect penis:
erection, (small penis, small testicles:1.1), cock sheath, dark testicles, pink penis"small" tag is highly unstable due to very small dataset
For large erect penis:
erection, large penis, large testicles, cock sheath, dark testicles, pink penis
For huge erect penis:
erection, huge penis, huge testicles, cock sheath, dark testicles, pink penis
For gigantic erect penis:
erection, gigantic penis, gigantic testicles, cock sheath, dark testicles, pink penis
If you're ran into any problems with this LoRA, you can always dm me on discord:
troubledarkness - discord nickaname
I'll be glad to help | Буду рад помочь.
Credits:
Trained on NAI Anime Full.
Dataset of 150 pictures (v1), 8 epochs, Clip Skip - 1
Dataset of 224 pictures (v2), 14 epochs, Clip Skip - 2
Dataset of 160 pictures (v3), 10 epochs, Clip Skip - 2
Dataset of 300 pictures (v4), 10 epochs, Clip Skip - 2
Dataset of 600 pictures (v5), 10 epochs, Clip Skip - 2
Dataset of 700 pictures (v6), 10 epochs, Clip Skip - 2
Dataset of 1000 pictures (v7), 20 epochs, Clip Skip - 2
Really big thanks to:
Some good folks from Discord server
You, for reading, downloading, liking and reviewing this LoRA
Description
What's new?
Extended training data up to 1000 images, cropped and properly tagged
For improved concept training retrained LoRA with 128/64 Dim/Alpha values
New model size is 148mb for better quality
Changed "big penis" tag to "small penis" for better training
Much better prompt guidance in general
Much better "from side" and "from behind" shots
Still can do penis growing from ass, but now decent quality is achievable
Much better color adjustment for penis and testicles
Better penis shape with less mistakes on state or curve
Still can do penis-less testicles, because it don't know about penis pointing towards viewer, there is not so much dataset about it
Better control over penis and testicles size, you can do smaller penis with bigger testicles, for example. At least, you have higher chanced for this.
use scheme like: (large penis:1.2), huge/gigantic testicles
Right now it's almost impossible for me to fully remove style impact for such big LoRA, so, i've tried to minimize this effect as possible.
Fixed frequent animal ears appears without prompting it
For most previous versions i've been using wrong training settings, which was leading to baking most things into single tag, that was averaging generation results. Now it's fixed.
FAQ
Comments (19)
What's new (v7)?
- Extended training data up to 1000 images, cropped and properly tagged
- For improved concept training retrained LoRA with 128/64 Dim/Alpha values
- New model size is 148mb for better quality
- Changed "big penis" tag to "small penis" for better training
- Much better prompt guidance in general
- Much better "from side" and "from behind" shots
-- Still can do penis growing from ass, but now decent quality is achievable
- Much better color adjustment for penis and testicles
- Better penis shape with less mistakes on state or curve
-- Still can do penis-less testicles, because it don't know about penis pointing towards viewer, there is not so much dataset about it
- Better control over penis and testicles size, you can do smaller penis with bigger testicles, for example. At least, you have higher chanced for this.
-- use scheme like: (large penis:1.2), huge/gigantic testicles
- Right now it's almost impossible for me to fully remove style impact for such big LoRA, so, i've tried to minimize this effect as possible.
- Fixed frequent animal ears appears without prompting it
For most previous versions i've been using wrong training settings, which was leading to baking most things into single tag, that was averaging generation results. Now it's fixed.
This is so much easier to use than other futa loras. Appreciate it!
@rayenearth i guess, because it was a long road of troubles, issues and fixes. There is 7 versions of this LoRA, and most futa LoRA barely has 2 versions. Also, they usually don't have large dataset, which is crucial for this type of LoRAs. As you know, basic SD or NAI / Anything models doesn't know much about penis, so, it needs a lot of training for this concepts to work at least on "good" level. Thanks for commentary and sharing your thoughts!
I adore this Lora since V3. Keep up the great work please!
No one else will ever give us horse penises like this..
Hi, thanks for keeping and eyes on this LoRA, there is a lot of challenged i'm facing in process, so, it gives me quite a lof of unique experience in LoRA training. I want to do regular penis LoRA too, but my time is very limited right now.
@TroubleDarkness It's great to hear that you're working on a regular penis LoRA!
@Pseudough yeah, there is not so much good penis LoRAs right now. But i'll need to stop working on my staff until 21 november, need to go to stationary hospital for some exams. But i'll continue to work on my models and LoRAs after that. Thanks for commentary!
@TroubleDarkness Good luck! No rush, real life is more important, this stuff is all for fun anyways.
If there is any assistance I can lend with the LoRA, I would love to help. I've got a better graphics card on the way and want to learn to make models.
beautiful lora, really customizable, but i have a question, will you make a normal futanari with the same customizability?
Hi! To make it clear, it's not a LoRA only for futanari. You can use horse penis LoRA with male characters as well. LoRA for regular penises will be the same customizable, yeah. But for the start, only basic tags for controls will be available, for example - sizes for both penis and testicles, state for testicles (sagging / resting) and all fluid-related tags. I think i need to start from 500 to 700 art in dataset for this LoRA. But first of all, i need to get home at least.
Do you have any advice for getting darker skin tones to work with this lora? It seems to really struggle with them - to the point that it will most often give the character dark arms and legs, but a pale torso.
Hi,
i've tried to generate a lot of characters before 7th version release, and it was working fine with dark-skinned characters (not tanned, sadly, there is almost no art with tanned characters with horse penises), I've tested it a little bit today, and now i can't get dark-skinned characters even with tag (dark skin, dark-skinned female/male) consistently.
The only one way i was able to get them - (dark-skinned female/male:1.4), (very dark skin:1.2) - in positives.
and
(pale skin:1.4) in negatives. But results are unstable still.
And i'm not sure why this happening. There is a lot pictures with dark skin tag in dataset.
Well. I've found a LoRA on CivitAI, that helps with generation of dark-skinned characters (for anime-based models at least):
-> https://civitai.com/models/81176/chocoskin
Tried it for 100 gens with different emphasis values, works very well. Not sure if this would work good with model you're using, but you can try.
I hope this would help. Thanks for pointing out that problem. I'll try to do something with this in the next update.
@TroubleDarkness I forgot to check this after posting; thank you for the response! I had actually noticed it back in... V6, I think? I'll give the Chocoskin lora a look. I wound up having some success with this one: https://civitai.com/models/112594/skin-tone-slider-lora - though it did insist on doing the 'swimsuit' tan sometimes.
I actually did have a question, though! The 'skin-colored testicles' tag - is that specifically for lighter-toned characters, or does it try to match to whatever the character's skin color is? It seems like the latter, but I'm not sure. Great LoRA all-around, though! Between this and its floppy variant, I've almost completely tossed aside 'regular' futanari LoRAs. The results with these are just consistently great.
@KSinclaire There is a problem with tag for light-skinned testicles, and even tag he have not working as intended. It just painting character skin to some colors because of word "skinned", blue or purple, usually. But it should do just light-skinned testicles, as you mentioned. And even if i'll try to add "light-skinned testicles" or just "light testicles" tag, this also would bring something related to lighting because of word "light" in prompt, so, i can't fix it. And right now i'm not able to find any good solution for this.
Adding something absolutely not related to color like "aqweqwe" tag for this color is a bad idea, i want this LoRA to be intuitive for most people.
Also, i've tried not to add tag at all, to make light-skinned testicles basic color, but this attempt was a failure. This color/concept was lost.
I've asked some creators about penis LoRAs, but most of them never did this kind of concept, and even those, who tried to do that, never did anything better that futa veiny LoRA or this horse penis LoRA. So, i need to investigate all problems on my own.
I hope i'll find a solution for that someday. And for other issues with this LoRA as well.
Thanks for replying!
Last time I tested this was with V3, do the newer versions work with SDXL at all?
Hi, for SDXL compatibility it needs to be trained for SDXL. All versions of this LoRA were made for SD 1.5 models. Also, right now there is model called "PonyDiffusionXL", that may do a lot of things, and also, horsecocks, so, such LoRAs are unnecessary.
Is an XL compatible version possible? This lora is wonderful but unfortunately it does not work on many XL models
It can't work on XL models because it was trained for SD 1.5 models. All content for both this versions are incompatible. Of course, the same for SD 1.5 / SD 2.0, SD 2.1 / SDXL models in general. There is also small incompatibility between SDXL models and PonyDiffusion-based models.
But i guess i'll not be able to gather 2000 images of horse penises again. So, i'm pretty sure there won't be an SDXL version of this LoRA. Especially knowing that there is already PonyDIffusion model, good at various penises types.
@TroubleDarkness It's kind of sad, this model is the best of its kind, if you're interested I can get you images of those penises haha












