Vibrant_Outline

UPDATES:

Updated with three new loras for Krea2, Anima, and Illustrious! These loras are an attempt to combine the old with the new. Using training data created from both of my Krea2 V1 and Illustrious V1 models, many more images were created for this purpose so hopefully this time around it will be easier to get more variety in your images!

Please note I was very excited to release the new V2 models and had been working on them for several weeks and multiple training sessions. So they have not yet been extensively tested. I'm confident that you'll get wonderful images, but what I have listed as defaults may not be entirely accurate, and I only shared what I was using at the time to create the sample images with each lora's posting. I encourage you to experiment and let me know what works for you! Please post your images and let me know how you like the lora and if you really enjoy something or if there's anything you find that doesn't seem to be working as intended. Happy generating!

Also, Thank you to @FUNLUCKY and @LawOfAI for allowing me to use several of their images to get more training data!

TRIGGER WORDS:

vibrant_outline - Base trigger word, will have a good chance of trigger either of the below triggers without using them, however using the triggers below will help to guide the image generation.

V2.0

multi_vibe - The same effect as multi_chrome from V1.0.

mono_vibe - The same effect as mono_chrome from V1.0.

*These were updated as mono_chrome was too close to monochrome and sometimes lead to prompt bleed. So in the interest of consistency both were changed to multi_vibe and mono_vibe respectively.

black - background / body / textures - Not trigger words per'se, but if you find the image is not cooperating with black colors, entering black background is usually enough to fix this issue, but sometimes more granular control with your entry might do you better. Leave a comment if you find something that works better.

V1.0 - (includes Krea2 V1.0)

multichrome - Will emphasize a multi-colored outline version of your image, you will have to guide it a bit but generally seems to work well. (EX: a brown outline of her leather armor, and a blonde outline of her long voluminous hair)

monochrome - Will emphasize a singular color, if you add additional colors like it may still provide that, but it will likely to be suppressed. You can use this to avoid color bleed if you want a particular color to be your primary color.

Recommend (This is what i used on the included images.)

Sampler: Euler + simple

Steps: 12 (using turbo checkpoint)

weights: 0.50 to 1.25

-After playing with the raw model abit, it seems the lora provides sharper thinner lines, but has a harder time with adherence.

Notes:

-[Applies to V1 models except Anima]

Due to the low number of images used to make this certain subjects associated with the training images may show up consistently incorrectly. (Such as prompting zombies makes skeletons). If you find this Lora not following your prompt lower the weight and it will usually correct itself.

-If your having a hard time with the model filling in areas with color, raise the weight and it should at the very least make it darker if not pure black.

-I have not tried much outside of these weights, so give it a test and it might work for you.

Recognition:

Thank you to:

@FUNLUCKY

@LawOfAI