Maya — Character LoRA

This is a character LoRA of an original character named Maya.

This is my first time training a LoRA, so I’d really appreciate any constructive feedback, criticism, or suggestions that could help me improve future versions.

The main goal of this LoRA was to maintain a consistent character identity across different prompts, outfits, poses, camera angles, expressions, and environments while still keeping enough flexibility for creative generations.

V1.2 Update:

Removed the recurring Japanese-themed background elements that appeared throughout the images and diversified the dataset to include a wider variety of environments and visual settings.

Recommended Settings

Trigger Word: Maya
LoRA Strength: 0.8–1.0
Best starting point: 0.9

You can experiment with lower strengths if you want more freedom from the base model, or increase the strength if you want stronger character consistency.

I recommend using an upscaler for the best image quality and sharper final results.

What I’d Love Feedback On

I’d especially appreciate feedback regarding:

• Facial consistency

• Character likeness

• Different camera angles

• Expressions

• Full-body generations

• Outfit flexibility

• Different lighting conditions

• Prompt responsiveness

• Overall image quality

If you test the LoRA, feel free to share your results and the settings you used.

Since this is my first character LoRA, any tips regarding training settings, dataset quality, captions, epochs, or general improvements would be greatly appreciated.

Thanks for checking it out!