This Model was trained on 197 Images of a work colleagues Pointillism/dot-art style.
This is the First LoRA that I've had a go at making, so It might not be the best.
It was trained for 100 epochs.
I wanted to see what I could do with an Art Style I'm familiar with.
Values Between 0.3 and 0.7 seem to work quite well. Adding words like Pointillism or dot drawing to the prompt seemed to be a bit overkill in my experiments.
I followed a guide to create the LoRA in Visions of Chaos
Description
I'm not sure what to put here, other than when uploading this, it asks about how many epochs and training steps. I know that the epochs was set to 100. The only reference I can see to training steps is referred to as warmup training steps. So if it is that what its asking for, then that was 200.
FAQ
Comments (2)
This can be a useful LoRA for monochrome images with anime checkpoints. You will often need to prompt for "monochrome" with many checkpoints if that is the look you want, otherwise you will get color. It works well with a wide array of styles and prompts. The effect is far more pronounced with anime/comic/cartoon checkpoints. Realistic checkpoints tend to add too much definition, making the image appear to be simply a black and white photograph rather than drawn art. A worthwhile tool to add to your collection. Five stars.
Cheers for Your feedback, I'll try some other checkpoints, and update the Images.
It's a more fitting description than mine.