Update app.py

app.py

@@ -1,19 +1,19 @@
 import streamlit as st
+from audiorecorder import audiorecorder
 from transformers import pipeline
 from diffusers import StableDiffusionPipeline
 import torch
-from PIL import Image
+from tempfile import NamedTemporaryFile
 
-# Title
 st.set_page_config(page_title="AI Meme Generator", page_icon="🎭")
 st.title("🎭 AI Meme Generator (Voice + Text)")
 
-# Load Whisper (speech-to-text)
+# Load Whisper
 @st.cache_resource
 def load_asr():
     return pipeline("automatic-speech-recognition", model="openai/whisper-tiny")
 
-# Load Stable Diffusion (text-to-image)
+# Load Stable Diffusion
 @st.cache_resource
 def load_sd():
     device = "cuda" if torch.cuda.is_available() else "cpu"
@@ -25,20 +25,13 @@ def load_sd():
 asr = load_asr()
 sd_pipe = load_sd()
 
-# Function: Voice -> Text
-def transcribe(audio_file):
-    result = asr(audio_file)
-    return result["text"]
-
-# Function: Text -> Meme
 def generate_meme(prompt):
-    image = sd_pipe(prompt).images[0]
-    return image
+    return sd_pipe(prompt).images[0]
 
-# Tabs for input type
+# Tabs
 tab1, tab2 = st.tabs(["📝 Text to Meme", "🎤 Voice to Meme"])
 
-# ---------------- Text to Meme ----------------
+# Text-to-Meme
 with tab1:
     text_input = st.text_area("Enter your meme idea")
     if st.button("Generate Meme", key="text_meme"):
@@ -49,15 +42,19 @@ with tab1:
         else:
             st.warning("Please enter some text!")
 
-# ---------------- Voice to Meme ----------------
+# Voice-to-Meme (with mic recording)
 with tab2:
-    audio_file = st.file_uploader("Upload your voice recording (mp3/wav)", type=["mp3", "wav"])
-    if st.button("Generate Meme from Voice", key="voice_meme"):
-        if audio_file is not None:
+    st.write("🎤 Record your voice below and create a meme!")
+    audio = audiorecorder("Click to record", "Click to stop recording")
+
+    if len(audio) > 0:
+        # Save temp audio file
+        with NamedTemporaryFile(suffix=".wav", delete=False) as f:
+            audio.export(f.name, format="wav")
+            text = asr(f.name)["text"]
+
+        st.success(f"Recognized Text: {text}")
+        if st.button("Generate Meme from Voice", key="voice_meme"):
             with st.spinner("Transcribing and generating meme..."):
-                text = transcribe(audio_file)
                 img = generate_meme(f"Meme style funny cartoon with text: {text}")
                 st.image(img, caption="Generated Meme")
-                st.success(f"Recognized Text: {text}")
-        else:
-            st.warning("Please upload a voice file!")