HarleyD (Character LoRA) – Pony / SDXL

세부 정보

모델 설명


HarleyD is a warm, grounded and expressive character LoRA trained for Pony Diffusion / SDXL.


She is defined by her natural presence, freckles, and soft, human expressiveness rather than exaggerated stylization. HarleyD works especially well in everyday scenes, outdoor walks, quiet
indoor moments, and character-focused storytelling.

This LoRA was carefully curated with a focus on:

clean identity learning

pose and expression variety

minimal style overfitting

It is not a style LoRA – it is a true character LoRA.

Character Highlights:

Warm, approachable facial features

Freckles as a stable identity marker

Natural body proportions

Strong pose generalization (standing, walking, sitting)

Works with both relaxed and expressive moods

Supports barefoot / footwear prompts reliably

HarleyD feels alive and human, not like a static “model pose”.

Training Details:
Base model: SDXL 1.0

Optimized for: Pony Diffusion (also compatible with other SDXL illustration models)

LoRA type: Character

Resolution: 1024×1024

Dataset: ~22 carefully curated images

Rank (dim): 16

Trigger word: harleyd

Captions were written manually with a Pony-friendly, low-noise tagging approach to avoid overfitting and preserve flexibility.

How to Use:
Trigger word (required):
harleyd

Recommended prompt structure:
harleyd, 1girl, [pose], [expression], [clothing], [environment]

Example prompts:
harleyd, walking outdoors, casual clothing, soft lighting

harleyd, sitting indoors, neutral expression, cozy atmosphere

harleyd, smiling, barefoot, summer day

You can explicitly control expressions using:

smile

closed mouth

open mouth

teeth

eyes closed

Notes & Tips:
This LoRA does not enforce a specific outfit or setting

Avoid stacking heavy style prompts – let the base model handle style

For best results, keep LoRA strength between 0.8 – 1.1

License & Usage:
Free for personal and creative use.
If you publish images using this LoRA, attribution is appreciated but not required.

이 모델로 만든 이미지