Image-generation_StableDiffusion_basics

"""
Sources:
https://www.youtube.com/watch?v=H4VGaC1Gs7o
https://www.youtube.com/watch?v=rgoxZe-eGic&t=119s

https://huggingface.co/stabilityai/stable-diffusion-2-1
pip install diffusers transformers accelerate scipy safetensors
pip install --upgrade diffusers
pip install --upgrade transformers

"""

import torch
#from transformers import SAFE_WEIGHTS_NAME

from diffusers import StableDiffusionPipeline, DPMSolverMultistepScheduler

model_id = "stabilityai/stable-diffusion-2-1"

# Use the DPMSolverMultistepScheduler (DPM-Solver++) scheduler here instead
pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16)
pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)
pipe.batch_size = 1 # para no acabar con la memoria que tenemos disponible
pipe = pipe.to("cuda")


""" 
prompt1 = "a photo of an astronaut riding a horse on mars"
image1 = pipe(prompt1).images[0]   
image1.save("astronaut_rides_horse.png")
"""

prompt = "a color photo of three horses in the countryside. In color. The style is as a Dali painting"
image = pipe(prompt).images[0] # selecciono el index 0 porque quiero la primera probabilidad
image.save("medieval_horses4_Dali.png")