Back to LoRA Models — Free AI Art Models for Flux, Illustrious & More

Monster In Me — Erotic Fantasy Horror [Flux, Illu]

Monster In Me — Erotic Fantasy Horror [Flux, Illu]
Category Concept
Quality 4.0/5
Maturity Hard

## Description

This is a versatile fantasy-horror themed LORA with detail enhancer focused on the topic of large monsters having sex with beautiful woman in strange places, often damp or public locations. This LORA includes a number of different sexual concepts and monstrous creatures, listed below. I will post smaller versions soon once I can test them for quality.

You can choose how explicit the images will be from simple nudity to full penetration and ejaculation. Controlling anal versus vaginal penetration is more difficult but it sometimes works. Hands and feet suffered a little bit, but you can still generate fully clothed people with the LORA enabled. Penises are usually attached to their owner, unless you request that they are floating.

All relevant keywords have been translated into l33t versions to avoid any existing and/or poisoned keywords. For best results, start your prompt with [monster] fucking woman, onarmor, [position] s3x, [position] penetration, fannypacking. The position can be anal or vaginal, oral is present but not reliable. You can provide an art style and mood, but the effect can be limited depending on how different they are from the training data. For example, this is a highly detailed, digital illustration in a fantasy art style, showcasing an intense, s3xual scene. can include more detail, but styles like pen and ink drawing will have a more limited (but still visible) effect on the image.

You should be able to control the skin and eye color of the monster, all features of the woman, and the color and texture of any ejaculate or fluids.

### Concepts

  • vaginal/anal penetration and insertion
  • detailed genitals, clitoris, labia, and vulva
  • cum and fluids
  • onarmor, fanny-packing, meatshield theme
  • size difference
  • dildos and some object insertion (hard to control)
  • stomach bulge (weak concept)
  • tentacles (weak concept)

### Monsters

Trained:

  • demon/balrog
  • skeleton
  • ogre/troll
  • vampire
  • zombie
  • werewolf
  • minotaur
  • lizard/dragon (weak)

Tested:

  • arachnid
  • cyborg
  • orc/goblin
  • squid/kraken
  • ghost (not transparent, needs another LORA for that)
  • serpent/lamia (frequently shown with arms and legs)
  • gnome/dwarf
  • giant mechanical pineapple (I did not think of this one, Llama did)

Please let me know if you find any other monsters and creatures that work.

## Download

### Monster In Me v3 alpha 1 Flux

Download monster-in-me-v3-alpha-1-flux.safetensors (76.31 MB)

SHA256: 0586af611a71ed45bf4df954002a5a55a54c8c4c01eb9daed09d43413c195fa8

Model
Base ModelFlux1Training Precisionfp8
LicenseNon-Commercial LicenseBase LicenseFlux1.Dev License
Network Type
NetworkLoraNetwork Precisionbf16
Training Parameters
OptimizerProdigySteps1533
D Coef4Training Images498
Dimension8Alpha8
Min SNR GammaNoneWeight Decay0.2

### Monster In Me v2 Merged

Download monster-in-me-v2-merged.safetensors (292.23 MB)

SHA256: 5d38221bb04dc9cdbb00f923387533cbc2951b631999a475264088f705537d7e

Model
Base ModelFlux1Training Precision
LicenseNon-Commercial LicenseBase LicenseFlux1.Dev License
Network Type
NetworkLoraNetwork Precision
Training Parameters
Dimension16Alpha16
Min SNR GammaNoneWeight DecayNone

### Monster In Me v2 Base

Download monster-in-me-v2-base.safetensors (74.91 MB)

SHA256: 673648e459f662f5899f1f91556f50e231879d376374df42e7532836f7f795c8

Model
Base ModelFlux1Training Precisionfp8
LicenseNon-Commercial LicenseBase LicenseFlux1.Dev License
Network Type
NetworkLoraNetwork Precisionbf16
Training Parameters
OptimizerProdigySteps2200
D Coef2Training Images702
Dimension8Alpha8
Min SNR GammaNoneWeight Decay0.1

### Monster In Me v1 Full

Download monster-in-me-v1-full.safetensors (1169.23 MB)

SHA256: 73001b1a7d8c1fafcc06a14c75fcbe51928097f515a3f544506fcd3a193640e4

Model
Base ModelFlux1Training Precisionfp8
LicenseNon-Commercial LicenseBase LicenseFlux1.Dev License
Network Type
NetworkLoraNetwork Precisionbf16
Training Parameters
Steps600
Learning Rate2e-04Training Images249
Dimension128Alpha128
Min SNR GammaNoneWeight DecayNone
  • < 0.5 for general detail enhancer and more cleavage
  • < 0.8 for stylized nudity usually without penetration
  • 1.0 for nudity and penetration
  • up to 1.2 for strong detail, cum, etc
  • 2.0 causes very poor results

A comparison chart is shown in https://civitai.com/images/25805747

I have tested this LORA in ComfyUI and Forge. Cannot comment on other tools or what strength they will need.

## Example Prompts

### Censored Keywords

  • penis: p3n1s
  • vagina: v4g1n4
  • nipple: n1ppl3
  • sex toy: d1ld0
  • sex: s3x
  • oral: 0r4l
  • breast: br34st
  • clitoral: cl1t0r4l
  • clitoris: cl1t0r1s
  • anus: 4nus
  • anal: 4n4l
  • ass: 4ss
  • buttocks: 4ss
  • butt: 4ss
  • dildo: d1ld0
  • naked: n4k3d
  • nude: nud3

### Other Words And Phrases

  • pregnant
  • vulva
  • small/large br34sts
  • erect
  • perky
  • ejaculating
  • shaved pubic area
  • insertion
  • penetration

## Dataset

The dataset had 249 images all of them at least 1280x1600 resolution. The LORA was trained at 1280 resolution, but it can generate from 768 to 1280 and perhaps more. Works well with the hires fix.

All of the training images were generated by me and all characters were shown as consenting adult humanoids.

Images were captioned using a modified version of https://huggingface.co/spaces/fancyfeast/joy-caption-pre-alpha with prompting to be more explicit. None of the captions contain girl or 1girl and I have not observed this LORA generate any children without being prompted.

## Credits

The images used to train this LORA were generated using a combination of the following checkpoints and other LORAs:

Along with a number of character LORAs, including:

Images were upscaled with https://openmodeldb.info/models/2x-LexicaSwinIR and detailed with:

I believe this license matches all of the source models, since this was trained on generated images, and is not a merge of any kind. If you have any concerns, please DM me.