I was planned to trained this lora a long time ago and while I was working on dataset the other similar lora has been released. Anyway I am releasing it, so now you have a choice.

1. Main action variations

There are few concept variations it was trained on:

1.1 Regular Facefuck

Tag "facefuck" in the beggining and the main action is described with the next phases:

A man is facefucking a woman

A man facefucks a woman

A man is fucking a woman's mouth

A man fucks a woman's mouth

A woman is being facefucked by a man

A woman's face being fucked by a man

1.2 Impaling

Tag "impaling" and the next phrases:

With his hips remaining stationary, a man fucks a woman's face, impaling her throat with his penis. 

A man stands still impaling a woman's face on his penis

A man forcefully drives woman's head up and down on his erect penis.

"A man thrusts woman's face into his crotch, making her take his penis down her throat.

1.3 Excessive pressure

Train files were martked with tag "excessive pressure" and with the next phrases:

The man pushes his penis into the woman's mouth as he holds her head firmly in place.

The man forces his penis deep into the woman's throat while keeping a strong hold on her head.

2 Insertion control

I have included "half-insertion" and "full-insertion" tags captioning corresponding samples. Also I have added natural text for that as well. For example

The woman takes the entire length of his penis down her throat.

The woman takes only half of the man's down her throat and can't take it deeper as it hits the back of her throats.

But here I was less systematic, so it not reliably works.

3 Framing control

To contol distance to the camera I have used tags "close-up", "medium shot", "full shot". And looks like full-shot does not works fine, because there was a lot fewer samples for that kind on shot.

4 Movements intensity

I have used the next tags for movements: "slow movements", "fast movements". Samples with average speed do not have any speed tag.

• "slow movements" tag I sometimes extended in natural languare with phrase "their movements are slow and deliberate".

• "fast movements" tag sometimes was extended with "his movements are fast and forceful"

5 Positioning

There are three main positions which I added to the train data. For each position we have a tag at the beggining of the caption with extended with natural language description

• On all fours
  tag: "on all fours"
  example description: "The woman is positions on all fours", "she is on all fours"

• On stomach
  tag: "on stomach"
  example description: "The woman is lying on her stomach an a couch/bed/table"

• On squats
  tag: "on squats"
  example description: "The woman is sitting on squats with her legs wide spread"

• On knees
  I didn't added keyword tag for kneeling, but it works with natural language description

6 Bonus

Here is a wildcard template I used for testing. It is not exectly follows the principles how I conbstructed the captions for training, but I am too lazy to write a lot of templates. Maybe I will write 3 templates for each action. But now I dont want to.

{medium shot|full shot}, {impaling|facefuck}, full-insertion
A man forces fully drives a woman's head pushing his penis into the woman's mouth. He holds her head firmly in place, while she is gagging taking the full length of his penis. The woman's {blonde|red|bunette} hair {tied back in a ponytail|is messy|cascading down her shoulders} and she is wearing a {black top and gold jewelry|nothing|white crop top|dress}, while the man, partially visible, is naked. {Their movements are slow and deliberate|their movements are fast and forceful}. Her {small|massive full|massive sagy|medium sized} breasts are {hidden under a lacy bra|and swinging from side to side|prominently displayed}. {Her mascara is smeared|she has huge cute eyes|she wears a leather choker}. She has {pale|fair|tanned} skin, while the man has {black|pale|fair} skin. The scene is set {in a dimly lit room against a concrete wall|in deep green forest|on a sandy beach, they bodies are shiny from sweat.}