eye concept SDXL and Pony v1

Eye Concept SDXL / Pony LoRA | Hyper-Realistic Iris Detail

The definitive eye-detail LoRA for SDXL 1.0 & Pony checkpoints.

Eye Concept is a detail specialist that adds ultra-high-resolution, photorealistic iris structure to human eyes: crisp limbal rings, internal fiber detail, and natural specular reflections/catchlights. It upgrades SDXL eyes into detailed focal points while remaining easy to blend with both photoreal SDXL and stylized Pony models.

Key Features

• Universal SDXL Compatibility: Works with SDXL 1.0 and popular Pony-based checkpoints (e.g., Pony Diffusion).

• Multi-Color Training: Supports multiple real-world eye colors (blue / green / brown / hazel) with consistent micro-detail.

• Hyper-Detail Focus (Low Dim): Trained with a low network rank (Dim: 8) to emphasize iris structure over broader style changes.

• Seamless Blending: Enhances eye detail without overwriting the base model’s overall aesthetic.

Usage (Quick Start)

Base Model: SDXL 1.0 or any Pony checkpoint
Trigger Word: eye_concept
Recommended Weight: 0.5 – 0.7 (start at 0.6)

Presets

Preset A — Portrait Safe (most use-cases)

• <lora:eye_concept:0.45–0.55>
  Best when you want subtle realism without “too sharp” eyes.

Preset B — Hero Close-Up / Macro

• <lora:eye_concept:0.6–0.75>
  Best for close-ups where the iris fills more of the frame.

Prompt Examples

SDXL Photoreal (clean & natural)

photorealistic portrait, <lora:eye_concept:0.6>, cinematic lighting, sharp iris texture, natural catchlight, detailed limbal ring

SDXL With Color Control

photorealistic portrait, <lora:eye_concept:0.6>, emerald green eyes, sharp iris texture, natural catchlight

Pony (stylized base, detailed eyes)

portrait, <lora:eye_concept:0.55>, detailed iris texture, natural catchlight, (green eyes:1.2)

Notes (SDXL + Pony)

• Pony models will keep their base aesthetic (stylized/anime), but eyes gain significantly more structure and realism.

• If eyes look too “strong” or too sharp, lower the weight (try 0.45–0.55).

• Avoid stacking multiple heavy “detail/sharpness/skin” LoRAs at high weights—eye detail can become overly crisp when combined.

Bonus Micro-Detail (Optional)

Depending on your prompt, this LoRA may also enhance micro skin detail (e.g., freckles / sun-kissed look) due to the close-up training focus.

Mini prompt:
close-up portrait, <lora:eye_concept:0.55>, sun-kissed skin, natural freckles, sharp iris texture, soft daylight