<h2 id="eye-concept-flux-lora-or-hyper-realistic-iris-detail">Eye Concept Flux LoRA | Hyper-Realistic Iris Detail</h2>The definitive eye-detail LoRA for Flux models (requires a Style LoRA in most cases).Eye Concept is a detail specialist that injects ultra-high-resolution, photorealistic iris structure into human eyes: crisp limbal rings, internal fiber detail, and natural specular reflections/catchlights. This LoRA is not a style LoRA. For consistent photoreal output, it should usually be paired with a style-defining realism/cinematic LoRA.<hr /><h2 id="key-features">Key Features</h2><ul><li>Universal Flux Compatibility: Works with Flux and Flux-compatible checkpoints.</li><li>Detail Specialist: Focused on iris texture, reflections, and internal fiber structure.</li><li>Style Control via Pairing: Pair with a realism/cinematic style LoRA to avoid comic/stylized drift.</li><li>Bonus Micro-Detail (optional): Can enhance skin micro-detail (freckles / sun-kissed look) when prompted.</li></ul><hr /><h2 id="usage-(recommended-setup)">Usage (Recommended Setup)</h2>Base Model: Flux or any compatible checkpoint Trigger Word: <code>eye_concept</code><h3 id="recommended-weights">Recommended Weights</h3><ul><li>Eye Concept Weight: 0.5 – 0.7 (start at 0.6)</li><li>Realism/Style LoRA (optional but recommended): 0.6 – 0.85</li></ul><hr /><h2 id="practical-presets-(how-i-use-it)">Practical Presets (how I use it)</h2><h3 id="preset-a-general-photoreal-(recommended)">Preset A — General Photoreal (recommended)</h3>Use a realism/style LoRA to control the overall look.<ul><li><code>&lt;lora:flux_realism_or_style:0.7&gt;</code></li><li><code>&lt;lora:eye_concept:0.55–0.65&gt;</code></li></ul>Example: <code>photorealistic portrait, &lt;lora:flux_realism_or_style:0.7&gt;, &lt;lora:eye_concept:0.6&gt;, cinematic lighting, sharp iris texture, natural catchlight</code><h3 id="preset-b-character-lora-mode-(often-works-without-realism-lora)">Preset B — Character LoRA Mode (often works without realism LoRA)</h3>When using a strong character LoRA, I often skip the realism LoRA.<ul><li><code>&lt;lora:CHARACTER:0.7–1.0&gt;</code></li><li><code>&lt;lora:eye_concept:0.35–0.55&gt;</code></li></ul>Example: <code>photorealistic portrait, &lt;lora:CHARACTER:0.85&gt;, &lt;lora:eye_concept:0.45&gt;, natural lighting, sharp iris texture, natural catchlight</code><h3 id="preset-c-hero-close-up">Preset C — Hero Close-Up</h3>For macro/close-ups where you want maximum iris detail.<ul><li><code>&lt;lora:flux_realism_or_style:0.65&gt;</code></li><li><code>&lt;lora:eye_concept:0.65–0.75&gt;</code></li></ul>Example: <code>macro close-up portrait, &lt;lora:flux_realism_or_style:0.65&gt;, &lt;lora:eye_concept:0.7&gt;, shallow depth of field, sharp iris texture, natural catchlight</code><hr /><h2 id="bonus-features-mini-prompt-(sun-kissed-skin-+-freckles)">Bonus Features Mini Prompt (Sun-kissed Skin + Freckles)</h2>Use this when you want the extra micro-skin detail effect:<code>close-up portrait, &lt;lora:flux_realism_or_style:0.7&gt;, &lt;lora:eye_concept:0.55&gt;, sun-kissed skin, natural freckles, realistic skin texture, soft daylight, sharp iris texture</code>Tip: Trigger it as plain text: <code>sun-kissed skin</code> (no markdown/asterisks).<hr /><h2 id="important-notes-(style-+-skin-loras)">Important Notes (style + skin LoRAs)</h2><ul><li>Why a style LoRA is recommended: Eye Concept is a detail LoRA. Without a style/realism LoRA, Flux may drift toward comic/stylized output depending on your base checkpoint and prompt.</li><li>About “realistic skin” LoRAs: I generally avoid stacking heavy realistic-skin LoRAs with <code>eye_concept</code>, because both enhance high-frequency detail and can become overly sharp when combined. If you do stack them, lower one of the weights (e.g., reduce <code>eye_concept</code> by ~0.1).</li></ul>

Files with SHA256: c567a0cce62ff2d6d5df65372f268d7892f7ebdde76d9868f2acbeaf5912ff4d