Sign In

Hyper Slider Suite

236
2.8k
31.5k
106
Updated: Apr 14, 2024
conceptfurryhyper
Verified:
SafeTensor
Type
LoRA
Stats
1,349
17,386
Reviews
Published
Mar 14, 2024
Base Model
SD 1.5
Training
Steps: 1,000
Usage Tips
Clip Skip: 1
Strength: 1.5
Hash
AutoV2
7D01F588E6
default creator card background decoration
Yolup's Avatar
Yolup

These Models are Controllable Body Sliders

This is a suite of models trained to alter the magnitude of specific body parts one at a time, and the idea is that they are able to be combined to get the proportions of the exact character you want, no matter the size.

Notice

not all of the models are complete yet, but I wanted to publish this and roll them out as I complete them, keep an eye out for each new model I post here!

Why?

I was asked to make a better version of my "weight slider" LECO because it was inadequate for fine-tuned control over the subject's proportions. While attempting to make a LECO for all hyper proportions at once I eventually decided that I needed a separate slider for each and every major body part that one might want to be hyper. so I began an epic quest to train many sliders for this task.

How?

I used the LECO training script by p1atdev on GitHub to train these models, each minor version number indicates an entirely separate model that I trained (e.g. Hyper_HipsV1.21 means it was my 22nd model while training the Hyper_Hips variant) and I used a specialized "Configurator" that I wrote to oversee the training and generate the config files, the prompt schedule, and other such things according to some research I've read.

Donations

Speaking of training, training models is expensive and I run training on my own private server, if you like what I do, consider supporting the development here!

https://ko-fi.com/yolup

Key Benefits

  1. Modular LoRAs mean you get to mix and match what you want when you want it

  2. Expressing each concept with a LoRA makes it more controllable than with prompts

Quirks

  1. Controlling for stylistic changes was very difficult, so most models have stylistic deviations