## Diana - Highly Consistent Photorealistic Character LoRA (Flux.1 Dev)
An original character LoRA meticulously trained for high geometric facial consistency, raw skin textures, and dynamic environment adaptability across Flux.1 Dev architectures. This model does not require external face fixing nodes for standard portraits or medium framing.
---
### 💡 Operational Recommendations:
* Trigger Word: Diana
* FaceDetailer: Not necessary for close-up or medium shots. Disable or bypass the node (Ctrl+B) to keep raw skin textures intact. Only use it for wide full-body compositions.
* Styling Tips: For best consistency, maintain Diana with black or dark hair and dark brown eyes. Using the keywords twin high braids yields exceptionally clean and distinct structural results.
---
### ⚠️ Crucial: How to Use Dual-Text Prompts in Flux
To get the absolute maximum performance out of this model, split your text prompts according to your encoders:
* Top Prompt Box (CLIP-L): Tag List Only
* Method: Start with Diana, followed by short, unweighted keyword tags separated by commas.
* Example: Diana, woman, cinematic photo, night, leather jacket, twin high braids, 35mm film grain.
* Bottom Prompt Box (T5XXL): Natural Language Story
* Method: Write a descriptive, natural paragraph explaining the full scene as if talking to a person. Put Diana naturally anywhere in the sentence.
* Example: A highly detailed cinematic photograph of Diana sitting in a dimly lit cafe at night. Street lights outside cast soft reflections across her dark leather jacket.