app.py

import streamlit as st
import torch
from diffusers import StableDiffusionPipeline
from PIL import Image
import io
import base64

# Page configuration
st.set_page_config(
    page_title="Text to Image Generator",
    page_icon="🎨",
    layout="wide",
)

# Custom CSS for better UI
st.markdown("""
<style>
    .main {
        background-color: #f5f7f9;
    }
    .stApp {
        max-width: 1200px;
        margin: 0 auto;
    }
    .css-18e3th9 {
        padding-top: 2rem;
    }
    .css-1d391kg {
        padding: 3rem 1rem;
    }
    .stButton>button {
        background-color: #4CAF50;
        color: white;
        padding: 10px 24px;
        border: none;
        border-radius: 4px;
        cursor: pointer;
        font-size: 16px;
        transition: all 0.3s;
    }
    .stButton>button:hover {
        background-color: #45a049;
        box-shadow: 0 4px 8px rgba(0,0,0,0.1);
    }
    .download-link {
        display: inline-block;
        margin-top: 10px;
        padding: 10px 15px;
        background-color: #3498db;
        color: white;
        text-decoration: none;
        border-radius: 4px;
        text-align: center;
    }
    h1 {
        color: #2C3E50;
        margin-bottom: 30px;
    }
    .stSlider {
        padding-top: 1rem;
        padding-bottom: 1rem;
    }
</style>
""", unsafe_allow_html=True)

# Header
st.title("✨ Text to Image Generator")
st.markdown("Turn your imagination into images using AI! 🖼️")

@st.cache_resource
def load_model():
    model_id = "runwayml/stable-diffusion-v1-5"
    pipe = StableDiffusionPipeline.from_pretrained(
        model_id, 
        torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32
    )
    if torch.cuda.is_available():
        pipe = pipe.to("cuda")
    return pipe

# Create columns for the layout
col1, col2 = st.columns([1, 1])

with col1:
    # Text input for the prompt
    prompt = st.text_area(
        "Enter your prompt",
        placeholder="A beautiful sunset over the ocean with palm trees in the foreground",
        height=100
    )

    # Advanced options in an expander
    with st.expander("Advanced Options"):
        num_inference_steps = st.slider(
            "Quality (Steps)", 
            min_value=1, 
            max_value=100, 
            value=50,
            help="Higher values produce more detailed images but take longer"
        )
        
        guidance_scale = st.slider(
            "Creativity", 
            min_value=1.0, 
            max_value=20.0, 
            value=7.5, 
            step=0.5,
            help="Higher values make the image more closely follow the prompt"
        )
        
        seed = st.number_input(
            "Random Seed", 
            min_value=-1, 
            max_value=2147483647, 
            value=-1,
            help="Set a specific seed for reproducible results, or -1 for random"
        )
        
        image_width = st.select_slider(
            "Image Width",
            options=[512, 576, 640, 704, 768],
            value=512
        )
        
        image_height = st.select_slider(
            "Image Height",
            options=[512, 576, 640, 704, 768],
            value=512
        )

    # Generate button
    generate_button = st.button("Generate Image")

# Function to get a download link
def get_image_download_link(img, filename="generated_image.png", text="Download Image"):
    buffered = io.BytesIO()
    img.save(buffered, format="PNG")
    img_str = base64.b64encode(buffered.getvalue()).decode()
    href = f'<a href="data:image/png;base64,{img_str}" download="{filename}" class="download-link">{text}</a>'
    return href

with col2:
    # Display the generated image and provide a download link
    st.markdown("### Generated Image")
    
    result_container = st.empty()
    
    if generate_button and prompt:
        try:
            with st.spinner("Generating your image..."):
                # Load the model
                pipe = load_model()
                
                # Set the seed for reproducibility
                generator = None
                if seed != -1:
                    generator = torch.Generator().manual_seed(seed)
                
                # Generate the image
                image = pipe(
                    prompt=prompt,
                    num_inference_steps=num_inference_steps,
                    guidance_scale=guidance_scale,
                    generator=generator,
                    width=image_width,
                    height=image_height
                ).images[0]
                
                # Display the image
                result_container.image(image, caption=prompt, use_column_width=True)
                
                # Provide a download link
                st.markdown(get_image_download_link(image), unsafe_allow_html=True)
                
                # Display the parameters used
                st.markdown("### Parameters Used")
                st.text(f"Steps: {num_inference_steps}")
                st.text(f"Guidance Scale: {guidance_scale}")
                st.text(f"Seed: {seed}")
                st.text(f"Dimensions: {image_width}x{image_height}")
                
        except Exception as e:
            result_container.error(f"An error occurred: {str(e)}")
    else:
        result_container.markdown("Your image will appear here after generation.")

# Add footer
st.markdown("---")
st.markdown("Powered by Hugging Face Diffusers and Streamlit")