This LoRA emulates the Mythic style found in Harem Heroes. Mythic Illustrious is a high-fantasy digital illustration style defined by radiant divine auras, and intensely vivid color palettes. Characters are rendered with hyper-beautiful features, luminous eyes, and ornate fantasy detailing, surrounded by swirling magical energy and cinematic lighting that emphasizes depth, contrast, and grandeur. The style blends celestial mysticism with modern premium illustration polish, producing artwork that feels powerful, elegant, and otherworldly. Enjoy!
Description
This update is a complete rework of the training dataset and captioning process to teach the visual language of the Harem Heroes Mythic artwork rather than specific characters or scenes.
What's New
Rebuilt the dataset using 342 curated Harem Heroes Mythic illustrations.
Activation token:
HHMythicILv2. (I accidentally used v2 instead of v3)Captioning was redesigned to emphasize cinematic composition, dynamic perspective, lighting, depth, and color grading instead of character-specific attributes.
Improved consistency across the dataset by reducing unnecessary subject and anatomy tags.
Better learning of the signature Harem Heroes Mythic aesthetic, including:
Dramatic cinematic framing
Dynamic camera angles
Rich, vibrant color palettes
Strong lighting and depth
Glossy, polished rendering
Training Changes
Improved caption quality and consistency
Optimized style-focused tagging
Refined training parameters for better stylistic generalization
Results
Version 3 produces significantly more cohesive outputs while preserving the recognizable Harem Heroes Mythic aesthetic. Compared to the previous release, images show improved composition, lighting, color harmony, and overall visual consistency, making the style easier to invoke across a wider variety of characters and prompts.
FAQ
Comments (3)
The v3 lora has v2 in the filename. OCD trrrriiiggurred AAAA
Discovered the trigger word is too. Ded.
Yea, that's my bad 😅
I realized after training the file it came out like that and had to put the trigger word/sequence up. But I will say that your output is super gorgeous!
@HaremHelix Hehe no worries. I probably wouldn't retrain just for a trigger word if that happened to me xD
Actually I normally stay in the 2.5D style so wasn't too sure where I wanted to go with the entry. CivChan rightly called me out on it ha! The Lora did shift the aesthetic nicely on faces and more towards toon like style, here at 0.75.
It seemed to struggle a bit more on small eyes compared to base WAI, but that's typical SDXL some inpainting patched that right up.