About Krea 2 Version:

Please Study the example images and prompts closely. This can do a wide range of effects, but it is dependent on using strong style guidance tags like "gradient colors", "detailed linework", "gradient color anime style", "blended and bleeding watercolor gradients with rich saturation" etc. Be very specific about colors- "depth of field background of fading gradient blue hues with violet highlights" and similar. Do try it at different strength and experiment, and "prompt enhance" from the default Krea 2 ComfyUI workflow or in the on site generator is recommended.

About Illustrious Version:

This is a style LoRA which aims to help you create eye-catching images easily by combining vivid gradient colors with bold black silhouettes for a dynamic, high-contrast look. It is geared towards nature scenery such as forests, fields, deserts and beaches, but also does moody indoor and abstract scenes nicely. It is a little bit unpredictable, you might need a few generations to get exactly what you want, but it offers some cool combinations of style if you give it a chance. Putting the trigger words for this followed by what you want most at the start of your prompt is a good start. Compatibility varies strongly by base model, I will link a few suggestions at the bottom here.

Main trigger words you should always use: Gr@dientW@lds, color grading, gradient colors

Helpful words: nature, high contrast, vines, flowers, dynamic angle

To make an OC character lora work with this you need to overpower that lora's opinion of color/lighting. Try making sure this lora's trigger words come at the start of your prompt, use the other lora at a low weight around 0.3, and use color-altering character tags like "silhouette, dark character outline, character silhouette, glowing eyes, white eyes" etc. I will post a few OC characters to this gallery as examples you can examine if desired.

The training data contains only images I generated myself- nothing was stolen from other artists or the web. There is ZERO explicit content in the training data. I included a very few featureless nudes and silhouettes (for example breasts without details, a lower body shape without intimate parts) to help reinforce what the model should do with skin tones on characters, but there is no actual explicit detail whatsoever. That said, the model is versatile and will attempt whatever your base model and other loras ask of it, so be please aware and prompt with care. Feel free to look at the prompts in my example images and posts to this gallery for ideas and a sense of what this is capable of.

Please do consider leaving a detailed review with an image post, it helps me out and lets me see the cool images you've made. Any comments or constructive criticism are always greatly appreciated!

Finally, this is dedicated to the master storyteller @EannDelacroix, who's incredible character lore articles inspire me more and more. For always sharing their worlds and encouraging me to create my own, I hope the worlds this style creates make them smile as well :)