Today i am pleased to share with you my Mirufu enhancer lora.

It's aimed to solve a simple problem, more precisely, it is aimed to help when generating Milfs, aka mature women, and bringing up some more extra functions that you can find in the version section.

Base Version

This version was trained on many images of mature women, mostly from manga style drawings and illustrations, to enhance the tag "mirufu" for sd 1.5.

Why would we need another lora for this?

Welp, the answer is simple. This is pretty versatile and can help in the generation of nsfw images.

More precisely, while most lora out here affect the whole "body_type" of the subject, this lora let you choose (Without too many efforts), what to affect (see tags to have a better understanding)

Activation tag:

mifuru            <- the actual activation tag  (It's weight can be changed to have much different results)

mature_woman      <- it's optional, but it stabilize the image

More tags to consider:

breast dimensions

Medium_Breasts
Large_Breasts
Huge_Breasts

Other breasts tag:
Cleavage
Nipples (there's also the "covered nipples" one)

Style:  (Those tags are totally optional, but they are heavely affected by the lora so keep it in mind)

Monochrome
grayscale
Lineart

Suggested Weight:

The lora perform best between 0.6 and 1
But you have to consider few things, the optimal weight is around 0.7 - 0.8; going above will enhance the "Mature" side of the image, but will also affect the luminosity of the image (not in a bad way, but you might not wanna a result like that)

Here are some test of the lora:

Test 1

The only changing in the prompt was the addendum of "mirufu" and "mature woman" tags

The images are (in order from left to right): base img, 0.7 strenght, and lastly full strenght (1)

As you can see it works "similarly" to a slider, but not really.

Test2

Testing the grayscale and lineart (+ halftone_texture)

When working with manga images you might wanna tweak the "halftone_texture" tag (if you use it)

Test 3 & 4

This test was made to see the flexibility when used with other "character loras"

as you can see it works pretty good (sometimes, you might need to lower the character lora weight, like in the 1st test, other times like in the second case there's no need to tweek anything)

Last tip that i can give you is: if your model support the kl-f8-anime.vae use that, cause it gives almost the same result of vae-ft-mse-840000-ema, but it add some detail (especially when working with jewelry, like nutial ring, or idk necklaces)

Please Support me if you like the model by sharing your work, or by tipping buzzs, so i can continue training, even with my hardware limits.