Update app.py

app.py

@@ -9,16 +9,13 @@ from waitress import serve
 
 app = Flask(__name__)
 
-# Use whisper-small for faster processing and better speed
 device = "cuda" if torch.cuda.is_available() else "cpu"
 asr_model = pipeline("automatic-speech-recognition", model="openai/whisper-small", device=0 if device == "cuda" else -1)
 
-# Function to generate audio prompts
 def generate_audio_prompt(text, filename):
     tts = gTTS(text=text, lang="en")
     tts.save(os.path.join("static", filename))
 
-# Generate required voice prompts
 prompts = {
     "welcome": "Welcome to Biryani Hub.",
     "ask_name": "Tell me your name.",
@@ -29,7 +26,6 @@ prompts = {
 for key, text in prompts.items():
     generate_audio_prompt(text, f"{key}.mp3")
 
-# Symbol mapping for proper recognition
 SYMBOL_MAPPING = {
     "at the rate": "@",
     "at": "@",
@@ -42,20 +38,18 @@ SYMBOL_MAPPING = {
     "space": " "
 }
 
-# Function to convert audio to WAV format
 def convert_to_wav(input_path, output_path):
     try:
         audio = AudioSegment.from_file(input_path)
-        audio = audio.set_frame_rate(16000).set_channels(1)  # Convert to 16kHz, mono
+        audio = audio.set_frame_rate(16000).set_channels(1)
         audio.export(output_path, format="wav")
     except Exception as e:
         raise Exception(f"Audio conversion failed: {str(e)}")
 
-# Function to check if audio contains actual speech
 def is_silent_audio(audio_path):
     audio = AudioSegment.from_wav(audio_path)
-    nonsilent_parts = detect_nonsilent(audio, min_silence_len=500, silence_thresh=audio.dBFS-16)  # Reduced silence duration
-    return len(nonsilent_parts) == 0  # If no speech detected
+    nonsilent_parts = detect_nonsilent(audio, min_silence_len=500, silence_thresh=audio.dBFS-16)
+    return len(nonsilent_parts) == 0
 
 @app.route("/")
 def index():
@@ -72,14 +66,11 @@ def transcribe():
     audio_file.save(input_audio_path)
 
     try:
-        # Convert to WAV
         convert_to_wav(input_audio_path, output_audio_path)
 
-        # Check for silence
         if is_silent_audio(output_audio_path):
             return jsonify({"error": "No speech detected. Please try again."}), 400
 
-        # Use Whisper ASR model for transcription
         result = asr_model(output_audio_path, generate_kwargs={"language": "en"})
         transcribed_text = result["text"].strip().capitalize()
 
@@ -87,6 +78,5 @@ def transcribe():
     except Exception as e:
         return jsonify({"error": f"Speech recognition error: {str(e)}"}), 500
 
-# Start Production Server
 if __name__ == "__main__":
     serve(app, host="0.0.0.0", port=7860)