Update app.py

app.py

@@ -68,7 +68,13 @@ def transcribe(apikey, upload, audio, choice1):
 
     tts.tts_to_file(chat_response, speaker_wav = upload, language="en", file_path="output.wav")
     
-    audio_in = "output.wav"
+    rate, data = wavfile.read("output.wav")
+    
+    reduced_noise = nr.reduce_noise(y=data, sr=rate, prop_decrease= 0.95, stationary=True)
+    #reduced_noise = nr.reduce_noise(y = data, sr=rate, prop_decrease= 0.85)
+    #reduced_noise = nr.reduce_noise(y = data, sr=rate, thresh_n_mult_nonstationary=2, stationary=False)
+
+    wavfile.write("audio1.wav", rate, reduced_noise)
 
     enhance_model = SpectralMaskEnhancement.from_hparams(
     source="speechbrain/metricgan-plus-voicebank",
@@ -77,21 +83,13 @@ def transcribe(apikey, upload, audio, choice1):
     )
 
     noisy = enhance_model.load_audio(
-    audio_in
+    "audio1.wav"
     ).unsqueeze(0)
 
     enhanced = enhance_model.enhance_batch(noisy, lengths=torch.tensor([1.]))
     torchaudio.save("enhanced.wav", enhanced.cpu(), 16000)
 
-    rate, data = wavfile.read("enhanced.wav")
-    
-    reduced_noise = nr.reduce_noise(y=data, sr=rate, prop_decrease= 0.95, stationary=True)
-    #reduced_noise = nr.reduce_noise(y = data, sr=rate, prop_decrease= 0.85)
-    #reduced_noise = nr.reduce_noise(y = data, sr=rate, thresh_n_mult_nonstationary=2, stationary=False)
-
-    wavfile.write("audio1.wav", rate, reduced_noise)
-    
-    return [result.text, chat_response, "audio1.wav"]
+    return [result.text, chat_response, "enhanced.wav"]
 
 output_1 = gr.Textbox(label="Speech to Text")
 output_2 = gr.Textbox(label="ChatGPT Output")