This LoRA is part of a batch from my personal collection that I've decided to release publicly.
Because I'm putting out many at once, these haven't gone through the same extensive testing I usually do before release - but they've all been in active use in my own workflow for a while.
A lot of sliders on Civitai target the same concepts, and I know how hard it can be to tell the well-made ones apart from the broken ones. What I can promise is that anything I release is a proper quality slider: it does what it says, it doesn't wreck your image, and it combines cleanly with others. I recently stacked over 20 of my own LoRAs on a single generation and the output still came out great - so piling several of mine together is a safe bet for most situations.
Feedback and bug reports are welcome in the comments.
Recommended range is what i use in the sample image (tiny text at the bottom of the images)
Description
FAQ
Comments (5)
It slightly alters the character’s facial features and increases the thickness of the thighs. Around 12 seems to be the threshold—going higher starts to look a bit unnatural. Still, it’s a great LoRA, and I look forward to more of your work.
If the butt is big, the thighs should also be big for it to look harmonious.
An excellent slider Lora. I have tried other people's slider Loras, but only yours works clearly without affecting other features of the image. Could you share specifically how you trained it?
I use ai-toolkit and the concept slider lora option.
I usually train LoRAs at "Linear Rank" 1, "Data Type" FP32, Quantization "-NONE".
That will be a 10MB LoRA with minimal side effects. Or you can train at BF16 and have a 5MB file, but they are pretty much identical.
The rest is like the tutorial from Ostris on youtube.
Some times i train 10 runs before i get a LoRA that im happy with. And the main "skill" invoved with slider-loras is to prompt good so there is no/minimal side effects.
I also do some fine-adjustments with a tool i have made my self where i disable some of the layers in the lora that have side-effects i dont want, as well as normalize the LoRA to have a larger range, since i dont like the 0.0 to 1.0 scale., but that is just to get the final 5%.
@Loraholic Thank you very much