app.py

import streamlit as st
from PIL import Image
from transformers import pipeline
from gtts import gTTS
import tempfile

# —––––––– Page config
st.set_page_config(page_title="Storyteller for Kids", layout="centered")
st.title("🖼️ ➡️ 📖 Interactive Storyteller")

# —––––––– Cache model loading
@st.cache_resource
def load_pipelines():
    # 1) Image captioning
    captioner = pipeline(
        "image-captioning",
        model="Salesforce/blip-image-captioning-base"
    )
    # 2) Story generation with Flan-T5
    storyteller = pipeline(
        "text2text-generation",
        model="google/flan-t5-base"
    )
    return captioner, storyteller

captioner, storyteller = load_pipelines()

# —––––––– Image upload
uploaded = st.file_uploader("Upload an image:", type=["jpg", "jpeg", "png"])
if uploaded:
    image = Image.open(uploaded).convert("RGB")
    st.image(image, caption="Your image", use_column_width=True)

    # —––––––– 1. Caption
    with st.spinner("🔍 Looking at the image..."):
        cap = captioner(image)[0]["generated_text"]
    st.markdown(f"**Caption:** {cap}")

    # —––––––– 2. Story generation
    prompt = (
        "Write a playful, 50–100 word story for 3–10 year-old children "
        f"based on this description:\n\n“{cap}”\n\nStory:"
    )
    with st.spinner("✍️ Writing a story..."):
        out = storyteller(
            prompt,
            max_length=200,
            do_sample=True,
            top_p=0.9,
            temperature=0.8,
            num_return_sequences=1
        )
        story = out[0]["generated_text"].strip()
    st.markdown("**Story:**")
    st.write(story)

    # —––––––– 3. Text-to-Speech
    with st.spinner("🔊 Converting to speech..."):
        tts = gTTS(story, lang="en")
        tmp = tempfile.NamedTemporaryFile(suffix=".mp3", delete=False)
        tts.write_to_fp(tmp)
        tmp.flush()
    st.audio(tmp.name, format="audio/mp3")