Update README.md

README.md

@@ -24,4 +24,47 @@ instance_prompt: >-
 license: creativeml-openrail-m
 ---
 
+# 🖐️ Overview
+
+## Trigger Word
+
+You can use this character by entering the prompt below.  
+If you want to change the character's outfit or certain details, you can modify the prompt as needed.
+
+```text
+backgwa, solo, pale blue hair, cat ears, golden eyes, navy beret, white shirt, red ribbon, oversized sleeves, blue jacket
+```
+
+## Preview
+
 <Gallery />
+
+---
+
+# 🔧 Training
+
+This LoRA character is a research-oriented project intended to protect artists' copyrights, art styles, training rights, and intellectual property.
+
+No human-made artwork was used as training data for this LoRA.  
+Instead, a synthetic dataset of 28 images was created using AI-generated images from GPT Image 2, and these images were used for training.
+
+To generate the initial images, the original character image was provided as an image input to the [`gemma-4-E4B-it`](https://huggingface.co/google/gemma-4-E4B-it) model running locally.  
+With predefined prompts and instructions, the model was used to output prompts capable of reconstructing the character from the provided reference image.
+
+After that, images were generated with GPT Image 2 using the constructed prompts.  
+When high-quality generated images were obtained, some of them were also used as additional references for generating further character reconstruction images.
+
+The generated images were labeled using GPT-5.5.
+
+A total of 28 synthetic images were used for training, and the LoRA was trained using the SDXL-based [`LUMIERE-Q`](https://huggingface.co/BackGwa/LUMIERE-Q) model.  
+The training was performed using the following parameters.
+
+|Parameter|Setting|
+|---|---|
+|Base Model|[`LUMIERE-Q`](https://huggingface.co/BackGwa/LUMIERE-Q)|
+|Dataset Size|`28 images`|
+|Image Source|`GPT Image 2`|
+|Labeling Model|`GPT-5.5`|
+|Epochs|`8`|
+|Repeats|`10`|
+|Resolution|`1024x1024`|