Update app.py

app.py

@@ -3,7 +3,7 @@
 import streamlit as st
 from PIL import Image
 from transformers import pipeline
-import pyttsx3
+from gtts import gTTS
 import tempfile
 
 # —––––––– Page config
@@ -13,52 +13,45 @@ st.title("🖼️ ➡️ 📖 Interactive Storyteller")
 # —––––––– Model loading + warm-up
 @st.cache_resource
 def load_pipelines():
-    # 1) Keep the original BLIP-base for captions
+    # 1) Original BLIP-base captioner
     captioner = pipeline(
         "image-to-text",
         model="Salesforce/blip-image-captioning-base",
-        device=0  # if you have GPU; use -1 for CPU-only
+        device=0  # set to -1 if CPU-only
     )
-    # 2) Switch to a lightweight story model
+    # 2) Lightweight GPT-Neo for stories
     storyteller = pipeline(
         "text-generation",
         model="EleutherAI/gpt-neo-125M",
         device=0
     )
 
-    # Warm up with a dummy run so first real call is fast
+    # Warm-up so first real request is fast
     dummy = Image.new("RGB", (384, 384), color=(128, 128, 128))
     captioner(dummy)
     storyteller("Hello", max_new_tokens=1)
 
     return captioner, storyteller
 
-# —––––––– Initialize local TTS (offline)
-@st.cache_resource
-def init_tts_engine():
-    engine = pyttsx3.init()
-    engine.setProperty("rate", 150)   # words per minute
-    return engine
-
 captioner, storyteller = load_pipelines()
-tts_engine = init_tts_engine()
 
-# —––––––– Main UI
+# —––––––– Image upload & processing
 uploaded = st.file_uploader("Upload an image:", type=["jpg", "jpeg", "png"])
 if uploaded:
-    # 1) Resize image to reduce BLIP load
+    # 1) Load + downsize for faster vision encoding
     image = Image.open(uploaded).convert("RGB")
     image = image.resize((384, 384), Image.LANCZOS)
     st.image(image, caption="Your image", use_container_width=True)
 
-    # 2) Caption
+    # 2) Caption step
     with st.spinner("🔍 Generating caption..."):
         cap = captioner(image)[0]["generated_text"].strip()
     st.markdown(f"**Caption:** {cap}")
 
-    # 3) Story (greedy = fastest)
+    # 3) Story generation (greedy for speed)
     prompt = (
-        f"Tell an 80–100 word fun story for 3–10 year-olds based on:\n\n“{cap}”\n\nStory:"
+        f"Write an 80–100 word playful story for 3–10 year-olds "
+        f"based on this description:\n\n“{cap}”\n\nStory:"
     )
     with st.spinner("✍️ Generating story..."):
         out = storyteller(
@@ -70,9 +63,11 @@ if uploaded:
     st.markdown("**Story:**")
     st.write(story)
 
-    # 4) TTS (local, no network)
+    # 4) Text-to-Speech via gTTS (network-based)
     with st.spinner("🔊 Converting to speech..."):
-        tmp = tempfile.NamedTemporaryFile(suffix=".wav", delete=False)
-        tts_engine.save_to_file(story, tmp.name)
-        tts_engine.runAndWait()
-    st.audio(tmp.name)
+        tts = gTTS(text=story, lang="en")
+        tmp = tempfile.NamedTemporaryFile(suffix=".mp3", delete=False)
+        tts.write_to_fp(tmp)
+        tmp.flush()
+    st.audio(tmp.name, format="audio/mp3")
+