---
license: creativeml-openrail-m
base_model: wavespeed-ai/wan-2.2-14b
tags:
- lora
- wan-2.2
- character
- photorealistic
- female
instance_prompt: l3dy Agata
---

# l3dy Agata - Wan 2.2 LoRA

This is a LoRA trained on the appearance of **l3dy Agata**. 
The model was trained using the Wan-2.2 architecture, providing high-fidelity facial features and natural skin textures.

## Model Description
- **Character:** l3dy Agata
- **Base Model:** Wan-2.2 (14B)
- **Training Data:** 30 high-quality portraits in various poses, lighting, and expressions.
- **Trigger Word:** `l3dy Agata`

## Usage Instructions

### 1. Dual LoRA Setup
Since this model was trained on Wan-2.2, it produces two files. For the best results, it is recommended to use both:
* **High Noise LoRA:** Handles structure and composition.
* **Low Noise LoRA:** Handles fine details and skin texture.

### 2. Recommended Settings
* **Trigger Word:** Always include `l3dy Agata` at the beginning of your prompt.
* **LoRA Weight:** `0.7 - 0.9` (Start with 0.8 for both).
* **CFG Scale:** `5.0 - 6.5`
* **Sampling Steps:** `25 - 40`

## Prompt Example
> `l3dy Agata, close-up portrait, looking at camera, soft cinematic lighting, high quality photorealistic, detailed eyes, messy hair.`

## Ethical Lead
This model is intended for creative and artistic use only. Do not use this model to generate NSFW or harmful content.