app.py

import streamlit as st
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
from gtts import gTTS
import io
from PIL import Image

# Install PyTorch
try:
    import torch
except ImportError:
    st.warning("PyTorch is not installed. Installing PyTorch...")
    import subprocess
    subprocess.run(["pip", "install", "torch"])
    st.success("PyTorch has been successfully installed!")
    import torch

# Load the image captioning model
caption_model = pipeline("image-to-text", model="unography/blip-large-long-cap")

# Load the GPT-2 model for story generation
#story_generator = pipeline("text-generation", model="gpt2")
story_generator = pipeline("text-generation", model="distilbert/distilgpt2")

def generate_caption(image):
    # Generate the caption for the uploaded image
    caption = caption_model(image)[0]["generated_text"]
    return caption

#def generate_story(caption):
    # Generate the story based on the caption using the GPT-2 model
    #prompt = f"Starting with 'Once upon a time', based on the image described as '{caption}', here is a short and interesting story for children aged 3-10. The story is positive and happy in tone, with added imagination:\n\n"
    #story = story_generator(prompt, max_length=500, num_return_sequences=1)[0]["generated_text"]
    
    # Extract the story text from the generated output
    #story = story.split("\n\n")[1].strip()
    
    #return story

#def generate_story(caption):
    # Generate the story based on the caption
    #story = story_generator(caption, max_length=200, num_return_sequences=1)[0]["generated_text"]
    #return story

def generate_story(caption):
    # Generate the story based on the caption using the GPT-2 model
    prompt = f"Once upon a time, based on the image described as '{caption}', here is a short, simple, and engaging story for children aged 3-10. The story should be easy to understand, use age-appropriate language, and convey a positive message. Focus on the main elements in the image and create a story that sparks their imagination:\n\n"
    story = story_generator(prompt, max_length=500, num_return_sequences=1)[0]["generated_text"]
    
    # Extract the story text from the generated output
    story = story.split("\n\n")[1].strip()
    
    # Post-process the story (example: remove inappropriate words)
    inappropriate_words = ["violence", "horror", "scary"]
    for word in inappropriate_words:
        story = story.replace(word, "")

    # Limit the story to approximately 100 words
    words = story.split()
    if len(words) > 100:
        story = " ".join(words[:100]) + "..."
    
    return story
    

def convert_to_audio(story):
    # Convert the story to audio using gTTS
    tts = gTTS(text=story, lang="en")
    audio_bytes = io.BytesIO()
    tts.write_to_fp(audio_bytes)
    audio_bytes.seek(0)
    return audio_bytes

def main():
    st.title("Storytelling Application")
    
    # File uploader for the image (restricted to JPG)
    uploaded_image = st.file_uploader("Upload an image", type=["jpg"])
    
    if uploaded_image is not None:
        # Convert the uploaded image to PIL image
        image = Image.open(uploaded_image)
        
        # Display the uploaded image
        st.image(image, caption="Uploaded Image", use_container_width=True)
        
        # Generate the caption for the image
        caption = generate_caption(image)
        st.subheader("Generated Caption:")
        st.write(caption)
        
        # Generate the story based on the caption using the GPT-2 model
        story = generate_story(caption)
        st.subheader("Generated Story:")
        st.write(story)
        
        # Convert the story to audio
        audio_bytes = convert_to_audio(story)
        
        # Display the audio player
        st.audio(audio_bytes, format="audio/mp3")

if __name__ == "__main__":
    main()