InstinctiveDiffuse

InstinctiveDiffuse is fine-tuned from StableDiffusionXL base model that carefully trained on 100,000 free-copyright images, and carefully DPO tuned.

the base model is stabilityai/stable-diffusion-xl-base-1.0 and the model has improved on being sharper in image generation with prompt tag of sharp.

Usage

the usage of the model is available via github of InstinctiveDiffuse. please also consider using these keep words to improve your prompts: best quality, ocean render, HD, --version 8.

Direct Use

import torch
from diffusers import StableDiffusionXLPipeline, EulerDiscreteScheduler

base = "Instinct-AI/InstinctiveDiffuse"
pipe = StableDiffusionXLPipeline.from_pretrained(base, torch_dtype=torch.float16)
pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing")
pipe.enable_model_cpu_offload()

pipe(
    "An astronaut riding a yellow horse on mars",
    num_inference_steps=8,
    guidance_scale=1
).images[0]