LoRA to make men hairy(ier) except it makes them look melted too :)
The images on each page don't correlate to the model I'm dumb and didn't realise they weren't shared ðŸ˜please forgive me
\/ IMPORTANT NOTE \/
There are three versions offered, all of them bad but slightly different:
-AlrightALT (Recommended): NEWâ„¢ It's the same as Alright but I nuked half the blocks. Now should no longer effect the style of the image, and can be used as a pseudo-slider. More dependant on "hairy" being in positive or negative prompts as a result. Probably the last release maybe.
- Alright: Like okay but better. Probably as good as it's going to get without me getting a better collection of hairy men images. No TE training like Okay either.
- Okay: A more standard, usable LoRA. The effects are there but they're less face-melty. Will still impact pose and introduce nudity. Has no text encoder training so you don't need to turn that down. Effects are less noticeable but doesn't immediately deepfry your gens.
- Okay_ALT: Like okay but I like it sliiightly better.
- Bad: Doesn't have as strong of a melting effect on the overall style of the image, but isn't as pronounced
- Worse: Middleground, degrades image style but not as noticeably, is mostly usable in txt2img
- Worst: Degrades the entire image quality, still usable if you img2img from it as a base
Overall I recommend lowering strength_clip for best results since I am ape brained.
Known issues
I'm aware it looks like garbage, destroys faces, reduces image quality etc. This is because I have no idea what I'm doing.
It also will push towards more tanned skin, I tried to remedy this in the dataset by including some gingers but :shrug:. The poses also trend towards more front facing angles, and the majority of the dataset was nudes so it also trends towards that.
How do I get results that look like the preview images?
img2img is a must until I stop being lazy and tag thing better/get a better dataset. I recommend grabbing ComfyUI and setting up samplers with the LoRA at different strengths (or one with one without), ending with a higher CFG to define the image out.
Feel free to message me on discord @fleshtexture if you'd like help
Description
FAQ
Comments (4)
Just wanted to give a heads up that your first gallery pic may be unintentional as it doesn't include this LoRA and has "hairy" in the negative prompt.
I wouldn't want anyone to potentially miss this LoRA. :)
Thanks very much for it!
Intentional to show the slider aspect, probably should've done it as a comparison thing though :s I'll try put one up later today with actual labelled axis sksksks
"Alright" without the ALT with 1.3 strength gives them cocks more realistic than models trained on cocks, love it!
what's the trigger word?
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.