Read gr.Audio input via filepath

app.py

@@ -26,13 +26,12 @@ def get_voice(voice):
     global voice_rec
     voice_rec = voice
 
-voice = gr.Audio()
+voice = gr.Audio(type='filepath')
 label = [gr.Label(), gr.Image()]
 
 def voice_to_image(voice):
 
-    sample_rate, audio_data = voice
-    audio_data = audio_data.astype(np.float32) # / 32767.0
+    audio_data, sample_rate = librosa.load(voice)
 
     # Generate the spectrogram using librosa
     spectrogram = librosa.stft(audio_data)