Update app.py

app.py

@@ -19,13 +19,24 @@ model, tokenizer = load_tts_model()
 user_text = st.text_input("Enter your text here:")
 
 # Generate Audio on Button Click
-if st.button("Generate Speech"):
-    if not user_text:
-        st.warning("Please enter some text.")
-    else:
-        inputs = tokenizer(user_text, return_tensors="pt")
-        with torch.no_grad():
+if st.button("Generate Speech"):    
+    if not user_text:        
+        st.warning("Please enter some text.")    
+    else:        
+        inputs = tokenizer(user_text, return_tensors="pt")        
+        with torch.no_grad():            
             output = model(**inputs).waveform
 
-        # Play the Audio Directly
-        st.audio(output[0].numpy()) 
+        # Specify sample rate (assuming it's the correct rate for the model)
+        sample_rate = 16000  # Or replace with the correct sample rate for 'facebook/mms-tts-eng'
+
+        # Optionally save to a temporary file (if needed)
+        sf.write("temp_audio.wav", output[0].numpy(), sample_rate)  
+
+        # Choose one of the following playback methods:
+
+        # Method 1: Play from temporary file
+        st.audio("temp_audio.wav")  
+
+        # Method 2: Play directly with sample rate 
+        st.audio(output[0].numpy(), sample_rate=sample_rate)