Bathany skin from Fortnite🧼
Suggested prompt:
BathanyFN, 3d,1girl, gloves, hat, solo, dark skin, black hair, under-eye patches, fingerless gloves, dark-skinned female, pink headwear, bandaid, pink coat, lips, head tilt, nose, beanie, standing, weapon
Recommended Settings
To get the most out of Version 2.0, I recommend the following configuration. This model has been tuned to be more responsive, so you may find you need less "prompt weight" than before.
Generation Parameters
LoRA Weight: 0.7 - 0.9 (Start at 0.7 for the best balance of style and flexibility).
Sampling Method: Euler a or DPM++ 2M Karras.
Sampling Steps: 25 - 35 steps.
CFG Scale: 5.0 - 7.5 (Lower CFG allows for more natural skin textures).
Prompting Tips
For Skin Detail: Use keywords like highly detailed skin, subtle skin texture, or realistic skin pores. Because of the upgrade, you don't need to "spam" these tags anymore.
For Poses: The model now understands spatial tags better. Try using from above, dynamic pose, kneeling, or perspective to test the new flexibility.
Negative Prompting: Keep it simple. Overloading the negative prompt can sometimes "flatten" the new skin details.
Recommended Negatives: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality.
Optimization Note
If you are using Illustrious XL, remember that this model responds very well to "Natural Language" mixed with tags. Instead of just sitting, try sitting on a wooden chair with legs crossed to see the improved understanding in action.
Description
Version 2.0: The Efficiency Overhaul
Version 2.0 represents a complete architectural refinement. This update prioritizes versatile prompting and high-fidelity output by streamlining the underlying training process.
Key Enhancements:
Superior Fidelity: Drastic improvements to sharpness, lighting, and compositional balance.
Hyper-Realistic Textures: Specialized tuning for skin micro-details.
Dynamic Versatility: Significant boost to pose and camera angle flexibility without model distortion.
Precision Prompting: Enhanced natural language processing allows for more accurate results with shorter, simpler prompts.
Technical Specs:
Optimized Architecture: Reduced file footprint for faster loading and broader compatibility.
Curated Dataset: Re-captioned and denoised training data for cleaner attribute recognition.
Recommended Weight: 0.7 - 0.9