<h2 id="model-overview-and-description"><strong>🌟 Model Overview &amp; Description</strong></h2><p>Trigger Word: </p><pre><code>Sintia</code></pre><p>Base Model: Trained on Krea (Works best when paired with Krea AI or Flux / SDXL bases depending on your generation workflow).</p><p>Dataset: 25 high-resolution, hand-curated images focused on Southeast Asian features, lighting, and realistic facial geometry.</p><h2 id="what-does-your-model-do">What does your model do?</h2><p>This LoRA introduces Sintia, a consistent Southeast Asian character model designed for high-fidelity portraiture, realistic lighting, and modern aesthetic styling. It captures subtle facial proportions, natural skin texture, and accurate hair and eye characteristics.</p><h3 id="what's-it-for">What's it for?</h3><p>It is built for creators looking to generate a distinct, highly consistent Southeast Asian female persona across various photographic settings, outfits, lighting conditions, and dynamic poses.</p><p></p><h3 id="what-is-the-model-good-at">What is the model good at?</h3><ul><li><p>Facial Consistency: Maintains identity fidelity using the Sintia trigger word even when altering prompts for outfit, background, or camera angles.</p></li><li><p>Photorealistic Features: Captures natural skin tones, lighting nuances, and realistic proportions.</p></li><li><p>Versatility: Responds well to simple style modifiers (e.g., streetwear, formal dress, soft indoor lighting, golden hour shot).</p></li></ul><h3 id="what-should-it-be-used-for">What should it be used for?</h3><ul><li><p>Photorealistic portraits and lifestyle photography renders.</p></li><li><p>Cinematic stills and fashion/outfit concepts.</p></li><li><p>Consistent character storytelling and visual projects.</p></li></ul><h3 id="what-is-your-resource-bad-at">What is your resource bad at?</h3><ul><li><p>Heavy Anime / Stylized 2D Art: Because the dataset consists of 25 curated realistic images, pushing it into extreme 2D anime, cel-shaded, or abstract art styles may bleed realistic facial details into the output.</p></li><li><p>Extremely Distant Shots: Like most small-dataset character LoRAs, facial details may degrade in extreme wide/background shots without using Adetailer or Hires.fix / Inpainting.</p></li></ul><h3 id="how-should-it-not-be-used">How should it not be used?</h3><ul><li><p>Do not use without the trigger word: Omitting Sintia will yield standard base-model faces.</p></li><li><p>Do not over-weight the LoRA: Running this model at weights above 1.0 (or 1.2) can cause oversaturation or artifacting. Keep weights between 0.6 and 0.9 for maximum flexibility.</p></li><li><p>Non-Consensual / Harmful Content: Please adhere to Civitai's Terms of Service regarding generated images.</p></li></ul><h2 id="recommended-generation-settings">⚙️ Recommended Generation Settings</h2><p>Trigger Word: Sintia</p><p>LoRA Weight: 0.7 – 0.85</p><p></p><p>Prompt Example:</p><pre><code>Sintia is wearing an espresso satin camisole with a low-rise ivory mini skirt. She sits sideways on a window bench, one leg crossed over the other while she turns her head back toward the camera. City lights blur softly behind her. The phone is slightly above eye level, giving a flattering downward angle without becoming a selfie cliché.</code></pre><pre><code></code></pre><p></p>

Files with SHA256: c1161368666801eda7e5731231dc199a958ffb1f451bb11c0a138d6071451ca8