Micromerge

New Model Name (e.g., ArtFusionXL)

This is a fine-tuned model based on VelvetToroyashi/WahtasticMerge.

Model Description

TIt has been trained on a dataset of approximately 15,000 images sourced primarily from ArtStation, X (k.a. Twitter), and OpenGameArt.

Training Data

The model was trained on a curated dataset of 15,000 images. The primary sources for these images were:

• ArtStation: For high-quality, professional digital art.
• X: For a diverse range of contemporary art styles.
• OpenGameArt: For assets related to game development, including characters and environments.

This diverse dataset aims to provide the model with a broad understanding of various artistic conventions and styles.

How to Use

This model can be used with any standard SDXL-compatible interface or library, e.g. Diffusers, Stable Diffusion WEBUI, ComfyUI.

Recommended Settings

For optimal results, we recommend the following inference parameters:

• Sampler: Euler or Euler ancestral
• Scheduler: Normal or Beta
• Steps: 16-24
• CFG Scale: 3-6
• Resolution: 832x1200 (or similar aspect ratios with a total area around 1024x1024)

Example Usage (Python with Diffusers)

from diffusers import AutoPipelineForText2Image
import torch

pipeline = AutoPipelineForText2Image.from_pretrained(
    "Pixel-Dust/Micromerge",
    torch_dtype=torch.float16,
    variant="fp16",
    use_safetensors=True
).to("cuda")

prompt = "a majestic fantasy landscape, vibrant colors, epic, detailed, masterpiece"
negative_prompt = "low quality, bad anatomy, deformed, ugly, distorted"

image = pipeline(
    prompt=prompt,
    negative_prompt=negative_prompt,
    num_inference_steps=20,
    guidance_scale=5,
    height=1200,
    width=832
).images

image.save("generated_image.png")