I'm thrilled to introduce this character LoRA, which I poured countless hours into perfecting on the LTX-2.0 model. This is my very first character LoRA for the LTX series, and I couldn't be more excited to share it with all of you.I've always been deeply captivated by the ethereal elegance and refined grace of Eastern women—their gentle gazes, fluid movements, and that dreamlike, luminous beauty when bathed in moonlight. "Yuexi" takes its name from the ancient Chinese poem "The moon rises bright, the lady is fair". She embodies everything I imagine in a timeless Eastern beauty, blending classical charm with modern versatility. My goal was to create a character with stunning photorealism and unlimited creative potential, one that can seamlessly fit into any scene, style, or story you envision.This is Yuexi LoRA V1.0. I'm already hard at work preparing the dataset for V2, which will be trained on LTX-2.3. The next version will include more diverse poses, full-body shots, varied lighting conditions, and an even wider range of clothing styles to make her even more flexible. Every piece of feedback you leave is incredibly valuable to me and will directly help me make the next iteration even better!## LoRA Core FeaturesThis model was trained on 100 carefully curated high-resolution images for a total of 48,000 steps, achieving the perfect balance between facial accuracy and creative freedom.✅ Precise Face Locking, Unlimited Creativity: Only facial features are fixed—body shape, hairstyle, clothing, makeup, environment, and actions are fully controllable via prompts✅ LTX Exclusive Optimization: Specifically addresses common flaws in LTX-2.3 native generation. distant shots, and high-dynamic scenes** like running, dancing, or turning—no more distorted or broken faces✅ Exceptional Generalization: Works flawlessly with ancient Chinese, modern, professional, athletic, wedding, and all other styles, performing reliably in any lighting or environment## Usage &amp; Recommended ParametersSimply load this LoRA after the LTX-2.0/2.3 base model.- Recommended Weight: 0.8-1.0 (start with 0.8 for optimal results)- Sampling Steps: 8 steps- CFG Scale: 1- Best Resolutions: 720x1280 / 1088×1920 (vertical), 1280x720 / 1920×1088 (horizontal)Thank you so much for your support! If you enjoy this LoRA, please consider leaving a like and a comment. It means the world to me and motivates me to keep creating better models for the community!

Files with SHA256: 54818e3e1fab825698f447d41c53a116b8730f104a7b5d2544af75d6e6f01075