DashAnimeXL V1 is a diffusion-based text-to-image generative model. The model is a finetune of SDXL by the research team at Dashtoon. Please see our blog for more details.
Model Description
- Developed by: Dashtoon
- Model type: Diffusion-based text-to-image generative model
- License: CreativeML Open RAIL++-M License
- Model Description: DashAnimeXL V1 is engineered to generate high-quality anime images from textual prompts. It features enhanced hand anatomy, better concept understanding, and prompt interpretation.
- Summary: This model generates images based on text prompts. It follows the same architecture as Stable Diffusion XL.
- Finetuned from model: SDXL
Using the model with 🧨 Diffusers
To use DashAnimeXL V1, install the required libraries as follows:
pip install diffusers --upgrade
pip install transformers accelerate safetensors
Example script for generating images with DashAnimeXL V1:
import torch
from diffusers import (
StableDiffusionXLPipeline,
EulerAncestralDiscreteScheduler,
AutoencoderKL
)
# Load VAE component
vae = AutoencoderKL.from_pretrained(
"madebyollin/sdxl-vae-fp16-fix",
torch_dtype=torch.bfloat16
)
# Configure the pipeline
pipe = StableDiffusionXLPipeline.from_pretrained(
"dashtoon/DashAnimeXL-V1",
vae=vae,
torch_dtype=torch.bfloat16,
use_safetensors=True,
)
pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config)
if torch.cuda.is_available():
pipe.to('cuda')
# Define prompts and generate image
prompt = "anime illustration, An ink painting with a superhot, pop art style, featuring vibrant splashes and gradient patterns merging with random signals and noise. A zoomed-in panda wearing glasses, appearing to look directly at the viewer. The piece is bathed in warm, volumetric lighting against a clear dusk sky background. The reflection in the panda's sunglasses reveals nuclear clouds, adding an element of surrealism."
negative_prompt = "nsfw, low quality, worst quality, very displeasing, 3d, watermark, signature, ugly, poorly drawn"
image = pipe(
prompt,
negative_prompt=negative_prompt,
width=1024,
height=1024,
guidance_scale=7,
num_inference_steps=20
).images[0]