JihyoSDXL – LoRA for SDXL

Base model: Stable Diffusion XL (SDXL 1.0/1.1 compatible)
Trigger word: jihyosdxl
Intended use: Safe-for-work image generation. Not intended for NSFW.

✨ Overview

jihyosdxl is a lightweight LoRA designed to enhance [insert subject/style focus, e.g., portrait realism, stylized characters, etc.] within SDXL while preserving natural anatomy and detail. It integrates smoothly into prompts—just include the trigger word jihyosdxl for activation.

• Optimized for SDXL’s higher fidelity

• Clean, balanced outputs at default settings

• Not designed for NSFW use

✅ Recommended Settings

• LoRA Weight: 1.0 (recommended)

  • Range: 0.7 – 1.2 to fine-tune style strength

• CFG Scale: 4.5 (sweet spot)

• Steps: More steps → more stable and detailed

  • Fast results: 30–40

  • High detail: 50–70

• Samplers: DPM++ 2M Karras, DPM++ SDE Karras

• Resolution: 832–1024px short side (for best SDXL results)

🧠 Prompting Tips

Example positive prompt:

jihyosdxl, a cinematic portrait of [subject], soft lighting, highly detailed textures, ultra realistic skin, 50mm lens, depth of field

General negative prompt:

blurry, low-res, distorted hands, extra limbs, watermark, text, signature, noisy background, artifacts

🚫 Content Policy

This model is not intended for NSFW generation. Please use responsibly in accordance with platform guidelines and local laws. With some base models it might trigger NSFW content.

📄 Notes

• At higher weights (>1.2), the LoRA may overpower SDXL’s default balance.

• Lowering CFG (to ~4.0) may help if results look too rigid or oversaturated.

• Increasing steps improves small detail stability.