app.py

import streamlit as st
from transformers import pipeline
import textwrap
import numpy as np
import soundfile as sf
import tempfile
import os
from PIL import Image

# Initialize pipelines with caching to avoid reloading
@st.cache_resource
def load_pipelines():
    # Load pipeline for generating captions from images
    captioner = pipeline("image-to-text", model="Salesforce/blip-image-captioning-large")
    # Load pipeline for generating stories from text prompts
    storyer = pipeline("text-generation", model="aspis/gpt2-genre-story-generation")
    # Load pipeline for converting text to speech
    tts = pipeline("text-to-speech", model="facebook/mms-tts-eng")
    return captioner, storyer, tts

# Load the pipelines once and reuse them
captioner, storyer, tts = load_pipelines()

# Function to generate caption, story, and audio from an uploaded image
def generate_content(image):
    # Convert the uploaded image to a PIL image format
    pil_image = Image.open(image)
    
    # Generate a caption based on the image content
    caption = captioner(pil_image)[0]["generated_text"]
    st.write("**🌟 What's in the picture: 🌟**")
    st.write(caption)

    # Create a prompt for generating a children's story
    prompt = (
        f"Write a funny, warm children's story for ages 3-10, 50–100 words, "
        f"in third-person narrative, that describes this scene exactly: {caption} "
        f"mention the exact place or venue within {caption}"
    )
    
    # Generate the story based on the prompt
    raw = storyer(
        prompt,
        max_new_tokens=150,
        temperature=0.7,
        top_p=0.9,
        no_repeat_ngram_size=2,
        return_full_text=False
    )[0]["generated_text"].strip()

    # Trim the generated story to a maximum of 100 words
    words = raw.split()
    story = " ".join(words[:100])
    st.write("**📖 Your funny story: 📖**")
    st.write(story)

    # Split the story into chunks of 200 characters for text-to-speech processing
    chunks = textwrap.wrap(story, width=200)
    
    # Generate and concatenate audio for each text chunk
    audio = np.concatenate([tts(chunk)["audio"].squeeze() for chunk in chunks])

    # Save the concatenated audio to a temporary WAV file
    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_file:
        sf.write(temp_file.name, audio, tts.model.config.sampling_rate)
        temp_file_path = temp_file.name

    return caption, story, temp_file_path

# Streamlit UI for the application
st.title("✨ Magic Story Maker ✨")
st.markdown("Upload a picture to make a funny story and hear it too! 📸")

# File uploader for image input
uploaded_image = st.file_uploader("Choose your picture", type=["jpg", "jpeg", "png"], help="Pick a photo to start the magic!")

# Placeholder image URL (replace with an actual URL of a child-friendly image)
placeholder_url = "https://example.com/placeholder_image.jpg"

if uploaded_image is None:
    st.image(placeholder_url, caption="Upload your picture here! 📷", use_column_width=True)
else:
    st.image(uploaded_image, caption="Your Picture 🌟", use_column_width=True)

if st.button("✨ Make My Story! ✨", help="Click to create your magic story"):
    if uploaded_image is not None:
        with st.spinner("🔮 Creating your magical story..."):
            caption, story, audio_path = generate_content(uploaded_image)
            st.success("🎉 Your story is ready! 🎉")
            st.audio(audio_path, format="audio/wav")
            os.remove(audio_path)
    else:
        st.warning("Please upload a picture first! 📸")