This model is a Low-Rank Adaptation (LoRA) fine-tuning of Animagine XL 4.0, specialized for generating high-quality anime-style images of the Jingliu character. The model has been trained on a curated dataset of 101 images featuring various poses, expressions, and scenarios.

Base Model

Base: cagliostrolab/animagine-xl-4.0
Architecture: Stable Diffusion XL
Resolution: 1024x1024

Training Details

Dataset

Dataset: Jingliu Character Dataset (101 images)
Character: Jingliu from Honkai: Star Rail
Image Types: Full body, upper body, portraits, action poses
Captions: Detailed danbooru-style tags

Usage

Option 1: Using LoRA Adapter (Recommended for flexibility)

python

import torch
from diffusers import StableDiffusionXLPipeline

# Load base model
pipeline = StableDiffusionXLPipeline.from_pretrained(
    "cagliostrolab/animagine-xl-4.0",
    torch_dtype=torch.float16
)
pipeline = pipeline.to("cuda")

# Load LoRA weights
pipeline.load_lora_weights("dddsaty/animagine-xl-jingliu-lora")

# Generate image
prompt = "jingliu, 1girl, solo, long hair, red eyes, holding, weapon, white hair, sword, dynamic pose, glowing eyes, blue theme"
negative_prompt = "nsfw, lowres, bad anatomy, bad hands, text, error, worst quality, low quality"

image = pipeline(
    prompt=prompt,
    negative_prompt=negative_prompt,
    num_inference_steps=30,
    guidance_scale=7.5,
    height=1024,
    width=1024
).images[0]

image.save("jingliu_output.png")

Option 2: Using Merged Model (Simpler, standalone)

python

import torch
from diffusers import StableDiffusionXLPipeline

# Load merged model directly
pipeline = StableDiffusionXLPipeline.from_pretrained(
    "dddsaty/animagine-xl-jingliu-lora",
    torch_dtype=torch.float16
)
pipeline = pipeline.to("cuda")

# Generate image
prompt = "jingliu, 1girl, solo, long hair, red eyes, holding, weapon, white hair, sword, dynamic pose, glowing eyes, blue theme"
negative_prompt = "nsfw, lowres, bad anatomy, bad hands, text, error, worst quality, low quality"

image = pipeline(
    prompt=prompt,
    negative_prompt=negative_prompt,
    num_inference_steps=30,
    guidance_scale=7.5,
).images[0]

image.save("jingliu_output.png")

Limitations

Optimized specifically for Jingliu character
May not generalize well to other characters
Best results with anime/illustration style
Requires trigger word "jingliu" for best results

License

This model inherits the license from the base model Animagine XL 4.0. Please refer to the original license for usage terms.

Citation

If you use this model, please credit:

bibtex

@misc{animagine-xl-jingliu-lora,
  author = {Your Name},
  title = {Animagine XL 4.0 - Jingliu LoRA},
  year = {2025},
  publisher = {HuggingFace},
  howpublished = {\url{https://huggingface.co/YOUR_USERNAME/animagine-xl-jingliu-lora}}
}

Base model:

bibtex

@misc{animagine-xl-4.0,
  author = {Cagliostro Research Lab},
  title = {Animagine XL 4.0},
  year = {2024},
  publisher = {HuggingFace},
  howpublished = {\url{https://huggingface.co/cagliostrolab/animagine-xl-4.0}}
}

Acknowledgments

Base model: Animagine XL 4.0 by Cagliostro Research Lab
Character: Jingliu from Honkai: Star Rail by HoYoverse
Training framework: Diffusers, PEFT