All the content I make is and will be completely free on CivitAI.
But if you want to support me with a coin, then you can do it here.
Sample images was generated with 🌕Luma
Recommended embeddings:
1. easynegative
2. bad-hands-5
3. badhandv4
Recommended extensions:
1. AutoChar
2. AfterDetailer
Strength 0.6-0.8 is optimal.
How to use:
Your prompt should be not too complicated. Shorter = better. Trigger words is 2dicon of for obects and props, 2dicon portrait of for portraits. Also i highly recommed to use this structure:wowicon of YOUR OBJECT, YOUR PROMPT
Sometimes require improvement:simple background, black background
Enjoy and share your feedback.
Description
FAQ
Comments (5)
Thanks for yet another great LORA. How does this one differ from your past ones for icons? I'm curious to know which one I should use in which scenario.
This one is trained on much bigger dataset (500 pics total) so it should be more flexible. But all dataset was with black background. On D.U.C.K. pack dataset was with inconsistent style so you may combine elements of all of them.
So this lora is kinda universal for objects and portraits. I even may say that this one should be basic cause it simple and very flexible. Close to RPG games icons.
D.U.C.K. is less consistent but may provide interesting and different results depends on your prompt. Close to Casual mobile games icons.
And Warcraft icons just Warcraft icons, square shape, glowing effects and stuff.
Great, thanks for the quick reply!
Thank you for the amazing LoRA
If you please bless me with your knowledge I am training a kohya dreambooth lora the guide said to resize the images which caused the character I am training to be stretched in all results so my question what should the person do in this case better crop the dataset or resize with white embed or something else? Also adding negative prompt stretched and wide shoulders fix the lora partially so is there a way to merge the negative prompt or even a checkpoint with a prompt so it is always triggered without the need to write?
I tried to find this info but no luck
Thank you in advance
Also is it better to train the lora on transparent pngs or to leave the character in scene without removing the background
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.