Modelpublic

dataautogpt3/CALAMITY

sourceHugging Faceapache-2.0updated 2y agoView on Hugging Face

8likes61downloads

Model Card

CALAMITY: Horror Themed Text-to-Image Model v1.0

CALAMITY is a specialized horror-themed text-to-image model built upon the Prometheus base model, which itself is a full rank finetune of the Playground v2.5 architecture. This checkpoint aims to generate haunting, surreal, and unsettling images with a distinctive aesthetic that sets it apart from mainstream models.

Key Features

Unique focus on dark, creepy, and horror themed visuals Trained on a curated dataset of horror imagery to capture genre-specific style Seamless integration with the Prometheus base model's enhanced accessibility features Advanced custom CLIP integration for improved text-to-image alignment Optimized for generating high-resolution, evocative horror artwork

Recommended Settings

Clip Skip: 2 CFG Scale: 7 Steps: 30-60 Sampler: DPM++ 2M SDE Scheduler: Karras Resolution: 1024x1024

Style Trigger word:"Sythentic Anime"

Use Cases

Generating evocative, nightmarish artwork for games, films, and books Concept art and visual brainstorming for horror-themed projects Exploring surreal, uncanny aesthetics and compositions Pushing creative boundaries in the horror genre through AI-augmented workflows

CALAMITY opens up new avenues for horror-themed image generation, providing creatives with a powerful tool for conjuring up unsettling and imaginative visuals. Experiment with different prompt combinations and settings to delve into a rich spectrum of macabre imagery.

Use it with 🧨 diffusers

python

import torch
from diffusers import (
    StableDiffusionXLPipeline, 
    KDPM2AncestralDiscreteScheduler,
    AutoencoderKL
)

# Load VAE component
vae = AutoencoderKL.from_pretrained(
    "madebyollin/sdxl-vae-fp16-fix", 
    torch_dtype=torch.float16
)

# Configure the pipeline
pipe = StableDiffusionXLPipeline.from_pretrained(
    "dataautogpt3/CALAMITY", 
    vae=vae,
    torch_dtype=torch.float16
)
pipe.scheduler = KDPM2AncestralDiscreteScheduler.from_config(pipe.scheduler.config)
pipe.to('cuda')

# Define prompts and generate image
prompt = "Sythentic Anime"
negative_prompt = ""

image = pipe(
    prompt, 
    negative_prompt=negative_prompt, 
    width=1024,
    height=1024,
    guidance_scale=7,
    num_inference_steps=50,
    clip_skip=2
).images[0]


image.save("generated_image.png")