Character LoRA - Maya v2.1

This LoRA was trained on Illustrious XL and features the character Maya with distinctive pink hair.

Trigger Words:

maya - Primary character trigger
pink_hair - Hair color/style trigger
long_hair - Hair length control (Added in v1.1)
ponytail - Optional hairstyle variation (Added in v1.1 - recommend "low ponytail" for original style)

What's New in v2.0:

Closest to original art style yet - Significant improvements over all v1.x versions
Enhanced preprocessing - Completely revamped training data preprocessing pipeline
Extended training cycles - Increased training steps with careful manual selection of optimal checkpoints
Hand-picked model - Each release candidate was manually evaluated for quality and sample consistency
Maintained compatibility - Tags and base model remain unchanged since v1.1 for seamless upgrading

What's New in v1.2:

Extended training with more epochs for improved stability
More consistent generation results
All v1.1 features remain compatible
The sample images may appear slightly worse than v1.1 because they are now generated using a much simpler workflow. For the best results, I recommend using waiNSFWIllustrious or other Illustrious models with sophisticated workflows like 2-pass generation.

What's New in v1.1:

Changed base model to waiNSFWIllustrious_v140.safetensors
Improved quality of training data
Standard generation methods should work better now
Art style now looks more like the original style
Added new trigger word "long_hair"
Optionally "ponytail" - if you prefer ponytail, recommend low ponytail to preserve original art style

Due to the original training images having poor quality, this LoRA may not work well with standard generation methods
Recommended workflow: Use the embedded ComfyUI workflow included in the sample images for best results
Load the workflow directly into ComfyUI for optimal generation quality