flux4 / app.py
salomonsky's picture
Update app.py
db53266 verified
from pathlib import Path
from PIL import Image
import streamlit as st
from huggingface_hub import InferenceClient, AsyncInferenceClient
import asyncio
import os
import random
import numpy as np
import yaml
try:
with open("config.yaml", "r") as file:
credentials = yaml.safe_load(file)
except Exception as e:
st.error(f"Error al cargar el archivo de configuración: {e}")
credentials = {"username": "", "password": ""}
MAX_SEED = np.iinfo(np.int32).max
client = AsyncInferenceClient()
llm_client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
DATA_PATH = Path("./data")
DATA_PATH.mkdir(exist_ok=True)
def authenticate_user(username, password):
return username == credentials["username"] and password == credentials["password"]
async def gen(prompts, width, height, model_name, num_variants=1, use_enhanced=True):
images = []
try:
for idx, prompt in enumerate(prompts):
seed = random.randint(0, MAX_SEED)
image, seed = await generate_image(prompt, width, height, seed, model_name)
image_path = save_image(image, f"generated_image_{seed}.jpg", prompt)
if image_path:
st.success(f"Imagen {idx + 1} generada")
images.append(str(image_path))
except Exception as e:
st.error(f"Error al generar imágenes: {e}")
return images
def list_saved_images():
return sorted(DATA_PATH.glob("*.jpg"), key=os.path.getmtime, reverse=True)
def save_prompt(prompt):
with open(DATA_PATH / "prompts.txt", "a") as f:
f.write(prompt + "\n")
def run_async(func, *args):
return asyncio.run(func(*args))
async def improve_prompt(prompt):
try:
instructions = [
"With this words, create a photorealistic description for a detailed txt2img prompt in English, 300 characters max.",
"With this idea, write a creative, realistic, and detailed text-to-image prompt in English, 300 characters max.",
"With this text, generate a descriptive and True to life txt2img prompt in English, 300 characters max.",
"With my idea, describe a photorealistic scene with detailed illumination for a txt2img prompt in English, 300 characters max.",
"With this concept, give a realistic, elegant txt2img prompt in English, emphasizing photorealism, 300 characters max.",
"With this perspective, conform a visually dynamic and hyperrealistic txt2img prompt in English, 300 characters max.",
"With this inspiration, realize a cinematic txt2img prompt in English with hyperrealistic elements, 300 characters max.",
"With my idea, make a lifelike and txt2img prompt in English, focusing on photorealistic depth, 300 characters max."
]
instruction = random.choice(instructions)
formatted_prompt = f"{prompt}: {instruction}"
response = llm_client.text_generation(formatted_prompt, max_new_tokens=100)
return response['generated_text'][:100] if 'generated_text' in response else response.strip()
except Exception as e:
return f"Error mejorando el prompt: {e}"
def save_image(image, file_name, prompt=None):
image_path = DATA_PATH / file_name
if image_path.exists():
st.warning(f"La imagen '{file_name}' ya existe en la galería. No se guardó.")
return None
else:
image.save(image_path, format="JPEG")
if prompt:
save_prompt(f"{file_name}: {prompt}")
return image_path
async def generate_image(prompt, width, height, seed, model_name):
if seed == -1:
seed = random.randint(0, MAX_SEED)
image = await client.text_to_image(prompt=prompt, height=height, width=width, model=model_name)
return image, seed
def get_prompt_for_image(image_name):
prompts = {}
try:
with open(DATA_PATH / "prompts.txt", "r") as f:
for line in f:
if line.startswith(image_name):
prompts[image_name] = line.split(": ", 1)[1].strip()
except FileNotFoundError:
return "No hay prompt asociado."
return prompts.get(image_name, "No hay prompt asociado.")
def login_form():
st.title("Iniciar Sesión")
username = st.text_input("Usuario", value="admin")
password = st.text_input("Contraseña", value="flux3x", type="password")
if st.button("Iniciar Sesión"):
if authenticate_user(username, password):
st.success("Autenticación exitosa.")
st.session_state['authenticated'] = True
else:
st.error("Credenciales incorrectas. Intenta de nuevo.")
async def generate_variations(prompt, num_variants, use_enhanced, style):
prompts = set()
while len(prompts) < num_variants:
if use_enhanced:
enhanced_prompt = await improve_prompt(f"{prompt}, estilo: {style}")
prompts.add(enhanced_prompt)
else:
prompts.add(f"{prompt}, estilo: {style}")
return list(prompts)
def image_viewer():
images = list_saved_images()
if images:
st.write("### Galería de Imágenes")
thumbnails = []
for image_file in images:
with Image.open(image_file) as img:
min_side = min(img.size)
cropped_img = img.crop(((img.width - min_side) // 2, (img.height - min_side) // 2,
(img.width + min_side) // 2, (img.height + min_side) // 2))
thumbnail = cropped_img.resize((100, 100))
thumbnails.append((image_file.name, thumbnail))
cols = st.columns(8)
for i, (name, thumbnail) in enumerate(thumbnails):
col = cols[i % 8]
with col:
st.image(thumbnail, use_column_width=False, width=100, caption=name)
if st.button("Ver", key=f"view_{i}"):
st.session_state['current_image_index'] = i
st.image(str(images[i]), caption=name)
else:
st.info("No hay imágenes guardadas.")
async def main():
st.set_page_config(layout="wide")
if 'authenticated' not in st.session_state or not st.session_state['authenticated']:
login_form()
return
prompt = st.sidebar.text_area("Descripción de la imagen", height=150, max_chars=900)
format_option = st.sidebar.selectbox("Formato", ["9:16", "16:9", "1:1"])
model_option = st.sidebar.selectbox("Modelo", ["black-forest-labs/FLUX.1-schnell", "black-forest-labs/FLUX.1-dev", "Shakker-Labs/FLUX.1-dev-LoRA-Logo-Design", "prithivMLmods/Logo-Design-Flux-LoRA"])
prompt_checkbox = st.sidebar.checkbox("Mejorar Prompt")
style_type = st.sidebar.selectbox("Estilo", ["Realismo", "Hiperrealismo", "Photorealismo", "Ilustración", "Abstracto", "Imaginativo"])
style_option = st.sidebar.selectbox("Camara", ["Full Shot", "Over Showder", "Medium Shot", "Long Shot", "Close-UP", "Dutch Angle", "High Angle", "Low Angle", "Oblique Angle"])
width, height = (360, 640) if format_option == "9:16" else (640, 360) if format_option == "16:9" else (640, 640)
if prompt_checkbox:
num_variants = st.sidebar.slider("Número de imágenes a generar", 1, 8, 1)
else:
num_variants = 1
if prompt_checkbox:
with st.spinner("Generando prompts mejorados..."):
prompts = await generate_variations(f"{prompt}, estilo: {style_option}, {style_type}", num_variants, prompt_checkbox, style_option) if prompt_checkbox else [f"{prompt}, estilo: {style_option}, {style_type}"]
else:
prompts = [f"{prompt}, estilo: {style_option}"]
if st.sidebar.button("Generar Imágenes"):
with st.spinner("Generando imágenes..."):
try:
results = await gen(prompts, width, height, model_option, num_variants, prompt_checkbox)
st.session_state['generated_image_paths'] = results
for result in results:
st.image(result, caption="Imagen Generada")
except Exception as e:
st.error(f"Error al generar las imágenes: {str(e)}")
image_viewer()
if __name__ == "__main__":
asyncio.run(main())