Update app.py

app.py

@@ -5,19 +5,21 @@ model = whisper.load_model("small")
 import torch
 import torchaudio
 from speechbrain.pretrained import SpectralMaskEnhancement
+from scipy.io import wavfile
+import noisereduce as nr
 import gradio as gr
 import openai
 
 mes1 = [
-    {"role": "system", "content": "You are a TOEFL examiner. Help me improve my oral Englsih and give me feedback. Replace the Arabic numerals with the corresponding English words in your response."}
+    {"role": "system", "content": "You are a TOEFL examiner. Help me improve my oral Englsih and give me feedback."}
 ]
 
 mes2 = [
-    {"role": "system", "content": "You are a mental health therapist. Your name is Tina. Replace the Arabic numerals with the corresponding English words in your response."}
+    {"role": "system", "content": "You are a mental health therapist. Your name is Tina."}
 ]
 
 mes3 = [
-    {"role": "system", "content": "You are my personal assistant. Your name is Alice. Replace the Arabic numerals with the corresponding English words in your response."}
+    {"role": "system", "content": "You are my personal assistant. Your name is Alice."}
 ]
 
 res = []
@@ -80,8 +82,16 @@ def transcribe(apikey, upload, audio, choice1):
 
     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, 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, "enhanced.wav"]
+    return [result.text, chat_response, "audio1.wav"]
 
 output_1 = gr.Textbox(label="Speech to Text")
 output_2 = gr.Textbox(label="ChatGPT Output")