Update app.py

app.py

@@ -21,7 +21,11 @@ if text:
     mel_outputs = tts_model.encode_batch(text)
     waveforms = hifi_gan.decode_batch(mel_outputs)
 
-    waveform =  waveforms[0].squeeze(1)
+    waveform =  waveforms[0].squeeze(1).numpy()
+
+    # Normalize the waveform to the range [-1, 1] if necessary
+    if np.max(np.abs(waveform)) > 1.0:
+        waveform /= np.max(np.abs(waveform))
 
     # Display the audio widget to play the synthesized speech
-    st.audio(waveform.numpy(), format="audio/wav", sample_rate = 22050)
+    st.audio(waveform, format="audio/wav", sample_rate = 22050)