Nilou_LoRa

Nilou_LoRa

---

Model description

base_model: circlestone-labs/Anima tags:

• lora
• genshin-impact
• nilou
• anime

---

Nilou LoRA (Anima-based)

This repository hosts a LoRA fine-tuned for Nilou from Genshin Impact. The model has been optimized for three distinct categories: Dancer Outfit, Casual Outfit, and NSFW.

Model Details

• Base Model: circlestone-labs/Anima
• Training Platform: Google Colab Modified version for pause/resume functionality
• Dataset Tagging: MohsinDev Dataset Tagger

Versions & Recommendations

> Note: For `Nilou-V2-E12`, a weight setting of 0.4 is recommended for the best balance of style fidelity and flexibility.

Usage

Trigger Words

Use the following trigger words based on your desired outfit category:

• Dancer Outfit: `nilou genshin, dancer outfit`
• Casual Outfit: `nilou genshin, casual outfit`
• NSFW: `nilou genshin, nsfw`

Training Credits

This LoRA was trained using the Anima LoRA Trainer App. The training script has been modified to support easy pause and resume functions during the training process.

---

Gallary

---

How to Train Your Own LoRA

Follow these steps to replicate the Nilou LoRA training process and create your own character models:

1. Prepare Your Dataset

• Collect Images: Gather high-quality, clear images from the web.
• Tagging: Use the MohsinDev Dataset Tagger to generate your .txt caption files.
• Consistency: Ensure each image has a corresponding text file with the exact same filename.

2. Training Environment

You can train using the Anima LoRA Trainer. Choose the version that best suits your needs:

• Modified Version support Pause and Resume (Recommended): Modified Colab Notebook — Features integrated pause/resume functionality for better session management.
• Original Version: Citron Legacy Anima Trainer — The original base implementation.

---

Download model

Download them in the Files & versions tab.

---

Test prompt

<Gallery />