app.py

import gradio as gr
from diffusers import StableDiffusionPipeline
import torch
import re

def split_into_sentences(text):
    """
    Splits the input text into individual sentences.
    This helps in identifying key scenes for image generation.
    """
    # Simple sentence splitter based on punctuation
    sentences = re.split(r'(?<=[.!?]) +', text)
    return sentences

def generate_comic_strip(story):
    """
    Generates a comic strip from the input story.
    
    Parameters:
    - story (str): The user's story prompt.
    
    Returns:
    - comic_strip (list): A list of generated images representing each scene.
    """
    if pipe is None:
        return ["https://via.placeholder.com/512x512.png?text=Model+Not+Loaded"]

    # Split the story into sentences to identify key scenes
    scenes = split_into_sentences(story)
    
    # Limit the number of scenes to prevent excessive image generation
    max_scenes = 3
    scenes = scenes[:max_scenes]
    
    comic_strip = []
    
    for idx, scene in enumerate(scenes):
        try:
            # Generate image for each scene with optimizations
            image = pipe(
                scene, 
                num_inference_steps=20,        # Reduced steps for faster generation
                height=256,                    # Reduced resolution
                width=256,                     # Reduced resolution
                guidance_scale=7.5,            # Default guidance scale
            ).images[0]
            comic_strip.append(image)
        except Exception as e:
            # In case of any error during image generation, append a placeholder image
            print(f"Error generating image for scene {idx+1}: {e}")
            comic_strip.append("https://via.placeholder.com/512x512.png?text=Image+Unavailable")
    
    return comic_strip

def main():
    """
    Sets up the Gradio interface for the GenArt Narrative application.
    """
    # Define the input component: A textbox for the user to input their story
    input_text = gr.Textbox(
        lines=5, 
        placeholder="Enter your short story here...", 
        label="Story Prompt"
    )
    
    # Define the output component: A gallery to display the generated comic strip
    output_gallery = gr.Gallery(
        label="Generated Comic Strip",
        columns=3,
        object_fit="contain",
        height="auto"
    )
    
    # Create the Gradio interface
    iface = gr.Interface(
        fn=generate_comic_strip,        # Function to process input and generate output
        inputs=input_text,              # Input component
        outputs=output_gallery,         # Output component
        title="GenArt Narrative",       # Title of the app
        description="Transform your short stories into engaging comic strips using AI-powered image generation.",  # Description
        examples=[                       # Example inputs for demonstration
            ["A young wizard discovers a hidden magical forest and befriends a talking owl."],
            ["An astronaut lands on a distant planet and encounters alien life forms."]
        ],
        allow_flagging="never",          # Disable flagging of outputs
        theme="default",                 # You can choose other themes like "huggingface"
    )
    
    # Launch the Gradio app
    iface.launch()

# Load the model globally to avoid reloading it for each request
pipe = None
try:
    print("Loading Stable Diffusion model...")
    # Initialize the Stable Diffusion pipeline with optimizations
    pipe = StableDiffusionPipeline.from_pretrained(
        "stabilityai/stable-diffusion-2-1-base",  # Model name
        torch_dtype=torch.float32,                # Use float32 for CPU
        low_cpu_mem_usage=True,                    # Optimize for CPU usage
        safety_checker=None,                       # Disable safety checker to speed up loading
        force_download=True                        # Force download to avoid resume_download warning
    )
    pipe = pipe.to("cpu")  # Move the model to CPU
    pipe.enable_attention_slicing()  # Reduce memory usage
    print("Model loaded successfully.")
except Exception as e:
    print(f"Error loading model: {e}")
    pipe = None

if __name__ == "__main__":
    if pipe is not None:
        main()
    else:
        print("Failed to load the model. Please check the error messages above.")