Version 1.5 coming "soon".

All samples are T2V generations and 100% fake. Check my profile for additional samples. This lora is not for the undressing of real people.

If you like my "work" please consider tipping my electricity bill: https://ko-fi.com/loralai82564

BTC: bc1qce69wlzt7mdw2q0rfj0afqrd3m73070yrg3uav

Upcoming loras: general purpose futanari/cowgirl (check my videos for a preview of that) and a good realistic cumshot lora (both for Wan 2.2 I2V) which are coming along nicely.

Instructions:

This is a lora for BBC blowjobs and handjobs, it can also do cock on head/face placement shots and maybe other body parts.

I2V & T2I untested. No trigger word.

Please check the sample videos I posted for some ideas of prompting. It's somewhat flexible, some examples of language to use:

"big black cock" ("penis" is not in the dataset captions at all, so don't use that)

"front_angle_shot" (to help trigger the front facing shot - check the sample with its usage... this is inconsistent and needs work, I will likely do a newer version of this lora soon)"

"a naked african american man walks into the scene/enters from the side"
"a naked african american man is sitting down in front of a woman/2 women..."

"she bobs her head up and down his big black cock"

"she is licking the sides of his big black cock"

"she sucks the tip of his big black cock"

"she licks the tip of his big black cock"
"she struggles to fit it in her mouth"

"gripping his big black cock at the base"

"she strokes his penis with her hand/hands"

"her lips stretch wide around his big black cock"

"he places his big black cock on her face, her face is partly obscured from view by his big black cock"

"she kisses his testicles"

"first person view/first person perspective/first person view camera"

"she sucks his testicles"
"viewed from the side"

"side shot"

There is sometimes some weirdness with random fingers appearing particularly at the testicles with some prompts.

Recommended settings:

Without speedup lora (I use 20-30 steps usually):

20 steps

Euler/dpmpp_2m (I usually use Euler)

Beta scheduler

CFG 3.5-5.0 for high/low models

9 steps on high noise, 11 steps on low noise with shift 5, thick bbc lora high strength 1.0, thick bbc lora low strength 0.75-1.0

11 steps on high noise, 9 steps on low noise with shift 8, , thick bbc lora high strength 1.0, thick bbc lora low strength 0.75-1.0

Lightx2v works pretty well with the low noise model, I generally do not recommend using the lightning lora for the high noise model.

With speedup lora:

20 steps

Euler/dpmpp_2m (I usually use Euler)

Beta scheduler

CFG 3.5-5.0 for high noise, 1.0 for low noise model

9 steps on high noise, 11 steps on low noise with shift 5, thick bbc lora high strength 1.0, thick bbc lora low strength 0.75-1.0

11 steps on high noise, 9 steps on low noise with shift 8, thick bbc lora high strength 1.0, thick bbc lora low strength 0.75-1.0

When using the speedup lora you can probably use less steps than 20 but that worked quite well for me, the problem with the speedup lora is it can cause excessive contrast and color and mess with the shape of the penis.

Negative prompt:

[default wan negative prompt] + hairy testicles, fingers randomly appearing (if they appear)

Training information:

Rank 32, alpha 1, Prodigy+Schedule Free, 4200 steps high, 4200 steps low, likely overtrained... a 2.0 version with a larger dataset is in the works.

Around 50 video clips and 25 images, 512x512 max for images, most videos 240x240 - 192x192

Upscale information:

I use Davinci Resolve to interpolate to 30fps and use its AI Super Scale for a bit of added clarity, I also add JPEG damage to make it look more like a real video. I prefer Resolve to using upscale models and RIFE in Comfy.