Home Artists Posts Import Register

Content

Hello Everyone!

Yesterday I've discovered this repo on Github:

ostris/ai-toolkit: Various AI scripts. Mostly Stable Diffusion stuff. (github.com)

In this repo there is a script named LoRA Slider Trainer, which is based off the work by p1atdev/LECO and rohitgandikota/erasing. By utilizing this script, one can effectively train a LoRA by specifying the actions to be taken when the weight is closer to -1 and those to be taken when it is closer to 1.

I've already started working on some experiments, in the pictures you'll find:

  • A background slider: -1 removes the background, while +1 improves the details
  • A color slider: -1 makes the picture monochrome, while +1 makes the picture more vibrant

These are not ready to be released yet because I am still trying things out. If you have any suggestions, please leave them in the comments below!

Files

Comments

No comments found for this post.