1mb LORA trained on three select vector illustrations with jagged trees\mountains and slightly 'melting' look, using luisa_pinguin guide.
3 ways to trigger, with slight differences in outputs:
Main trigger: vectorstyle.
No trigger, just add LORA to prompt.
Secondary, a bit weaker trigger,: limited vector.
Refer to examples to see possible results, feel free to experiment. Works good in combination with my previous style LORA.
If you liked my work, you can donate here.
Trained on Anything V4.5 with anime.vae.pt and CLIP Skip 2.
Samples are 512x768\768x512, using OrangeCocoa 50/50 Mix , with OrangeMix VAE, and using EasyNegative embedding, CLIP Skip 2.
Description
FAQ
Comments (8)
how you make your model weight so low? mine are always 140mb
I linked relevant guide in the description, short answer - set dim and alpha settings to 1
There is an interesting guide to optimice your LoRAs here: https://civitai.com/models/22530
Have you experimented with negative settings?
Seems that minimalist style LORAS can actually introduce MORE details into a picture at negative settings
I saw a post about it on subreddit, but didn't had time to test it out myself, only so much time in a day. But It's definitely on my testing list, I think playing with color or shape LORA's might be fun in that regard
What checkpoint model do you use?
OrangeCocoa 50/50 Mix , it's in the description
Would you consider making this for FLUX?
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.