app.py

import streamlit as st
import torch
from PIL import Image
import random
from diffusers import StableDiffusionInstructPix2PixPipeline, DiffusionPipeline, FluxPipeline

device = "cuda" if torch.cuda.is_available() else "cpu"

model_options = {
    "Instruct Pix2Pix": "timbrooks/instruct-pix2pix",
    "SDXL Turbo": "stabilityai/sdxl-turbo",
    "FLUX": "alimama-creative/FLUX.1-Turbo-Alpha"
}

selected_model = st.sidebar.selectbox("Selecciona el modelo", list(model_options.keys()))
model_id = model_options[selected_model]

@st.cache_resource
def load_model():
    if selected_model == "Instruct Pix2Pix":
        return StableDiffusionInstructPix2PixPipeline.from_pretrained(model_id, safety_checker=None).to(device)
    elif selected_model == "SDXL Turbo":
        return DiffusionPipeline.from_pretrained(model_id, safety_checker=None).to(device)
    else:
        return FluxPipeline.from_pretrained(model_id, safety_checker=None).to(device)

pipe = load_model()

def resize(img, max_size=512):
    ratio = min(max_size / img.width, max_size / img.height)
    new_size = (int(img.width * ratio), int(img.height * ratio))
    return img.resize(new_size, Image.LANCZOS)

def infer(source_img, prompt, steps, seed, guidance_scale):
    torch.manual_seed(seed)
    source_image = resize(source_img)
    progress_bar = st.progress(0)
    st.text("Generando imagen...")

    try:
        for i in range(steps):
            output = pipe(
                prompt,
                image=source_image,
                guidance_scale=guidance_scale / 10,  
                num_inference_steps=steps
            )

            result = output.images[0] if output.images else None

            if result is None:
                raise ValueError("No se generaron imágenes.")

            progress_bar.progress((i + 1) / steps)

        progress_bar.progress(1.0)
    except Exception as e:
        st.error(f"Error durante la inferencia: {str(e)}")
        return None

    return result

st.title("Flux Image to Image")

with st.sidebar:
    uploaded_image = st.file_uploader("Sube una imagen", type=["png", "jpg", "jpeg"], key="unique_file_uploader")
    prompt = st.text_input("Texto del prompt (máx. 77 tokens)")
    steps = st.slider("Número de Iteraciones", min_value=1, max_value=50, value=2, step=1)
    randomize_seed = st.radio("Randomize Seed", ["Randomize Seed", "Fix Seed"])
    seed = st.slider("Seed", min_value=0, max_value=9999, step=1, value=random.randint(0, 9999) if randomize_seed == "Randomize Seed" else 1)
    guidance_scale = st.slider("Guidance Scale", min_value=0.0, max_value=10.0, step=0.01, value=9.0)

if uploaded_image is not None and st.button("Generar imagen"):
    image = Image.open(uploaded_image).convert("RGB")
    result_image = infer(image, prompt, steps, seed, guidance_scale)
    if result_image is not None:
        st.image(result_image, caption="Imagen generada", use_column_width=True)