Update app.py

app.py

@@ -1,22 +1,20 @@
+import torch
 from flask import Flask, render_template, request, jsonify
 import os
-import torchafrom flask import Flask, render_template, request, jsonify
-import os
-import torch
 import re
-import ffmpeg  # Ensure FFmpeg is installed
+import ffmpeg
 from transformers import pipeline
 from gtts import gTTS
 from pydub import AudioSegment
 from pydub.silence import detect_nonsilent
 from waitress import serve
-import whisper  # Improved Whisper ASR Model
+import whisper  # Corrected whisper import
 
 app = Flask(__name__)
 
 # Load Whisper Model for Highly Accurate Speech-to-Text
 device = "cuda" if torch.cuda.is_available() else "cpu"
-asr_model = whisper.load_model("large-v3", device=device)
+asr_model = whisper.load_model("large-v3")
 
 # Function to generate audio prompts
 def generate_audio_prompt(text, filename):
@@ -47,7 +45,7 @@ SYMBOL_MAPPING = {
     "space": " "
 }
 
-# Function to convert audio to WAV format (Fixes FFmpeg issues)
+# Function to convert audio to WAV format
 def convert_to_wav(input_path, output_path):
     try:
         audio = AudioSegment.from_file(input_path)
@@ -55,7 +53,7 @@ def convert_to_wav(input_path, output_path):
     except Exception as e:
         raise Exception(f"Audio conversion failed: {str(e)}")
 
-# Function to clean transcribed text (Removes unnecessary words)
+# Function to clean transcribed text
 def clean_transcription(text):
     text = text.lower().strip()
     ignore_phrases = ["my name is", "this is", "i am", "it's", "name"]
@@ -67,11 +65,11 @@ def clean_transcription(text):
     
     return text.capitalize()
 
-# Function to check if the audio contains actual speech
+# 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)
-    return len(nonsilent_parts) == 0  # Returns True if silence detected
+    return len(nonsilent_parts) == 0
 
 @app.route("/")
 def index():
@@ -83,130 +81,26 @@ def transcribe():
         return jsonify({"error": "No audio file provided"}), 400
 
     audio_file = request.files["audio"]
-    input_audio_path = os.path.join("static", "temp_input")
+    input_audio_path = os.path.join("static", "temp_input.wav")
     output_audio_path = os.path.join("static", "temp.wav")
     audio_file.save(input_audio_path)
 
     try:
-        # Convert audio to WAV format
+        # Convert to WAV
         convert_to_wav(input_audio_path, output_audio_path)
 
-        # Check if the audio contains valid speech
+        # Check for silence
         if is_silent_audio(output_audio_path):
             return jsonify({"error": "No speech detected. Please try again."}), 400
         
         # Transcribe using Whisper
-        result = asr_model.transcribe(output_audio_path, language="en")
+        result = asr_model.transcribe(output_audio_path)
         transcribed_text = clean_transcription(result["text"])
         
         return jsonify({"text": transcribed_text})
     except Exception as e:
         return jsonify({"error": f"Speech recognition error: {str(e)}"}), 500
 
-# Use Waitress for Production Server
+# Start Waitress Production Server
 if __name__ == "__main__":
     serve(app, host="0.0.0.0", port=7860)
-
-import re
-import ffmpeg  # Ensures FFmpeg is installed
-from transformers import pipeline
-from gtts import gTTS
-from pydub import AudioSegment
-from pydub.silence import detect_nonsilent
-from waitress import serve
-import whisper_timestamped  # Improved Whisper with timestamps
-
-app = Flask(__name__)
-
-# Load Whisper Model for Highly Accurate Speech-to-Text
-device = "cuda" if torch.cuda.is_available() else "cpu"
-asr_model = whisper_timestamped.load_model("medium", device=device)
-
-# 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.",
-    "ask_email": "Please provide your email address.",
-    "thank_you": "Thank you for registration."
-}
-
-for key, text in prompts.items():
-    generate_audio_prompt(text, f"{key}.mp3")
-
-# Symbol mapping for proper recognition
-SYMBOL_MAPPING = {
-    "at the rate": "@",
-    "at": "@",
-    "dot": ".",
-    "underscore": "_",
-    "hash": "#",
-    "plus": "+",
-    "dash": "-",
-    "comma": ",",
-    "space": " "
-}
-
-# Function to convert audio to WAV format (Fixes FFmpeg issues)
-def convert_to_wav(input_path, output_path):
-    try:
-        audio = AudioSegment.from_file(input_path)
-        audio.export(output_path, format="wav")
-    except Exception as e:
-        raise Exception(f"Audio conversion failed: {str(e)}")
-
-# Function to clean transcribed text (Removes unnecessary words)
-def clean_transcription(text):
-    text = text.lower().strip()
-    ignore_phrases = ["my name is", "this is", "i am", "it's", "name"]
-    for phrase in ignore_phrases:
-        text = text.replace(phrase, "").strip()
-    
-    for word, symbol in SYMBOL_MAPPING.items():
-        text = text.replace(word, symbol)
-    
-    return text.capitalize()
-
-# Function to check if the 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)
-    return len(nonsilent_parts) == 0  # Returns True if silence detected
-
-@app.route("/")
-def index():
-    return render_template("index.html")
-
-@app.route("/transcribe", methods=["POST"])
-def transcribe():
-    if "audio" not in request.files:
-        return jsonify({"error": "No audio file provided"}), 400
-
-    audio_file = request.files["audio"]
-    input_audio_path = os.path.join("static", "temp_input")
-    output_audio_path = os.path.join("static", "temp.wav")
-    audio_file.save(input_audio_path)
-
-    try:
-        # Convert audio to WAV format
-        convert_to_wav(input_audio_path, output_audio_path)
-
-        # Check if the audio contains valid speech
-        if is_silent_audio(output_audio_path):
-            return jsonify({"error": "No speech detected. Please try again."}), 400
-        
-        # Transcribe using Whisper
-        result = asr_model.transcribe(output_audio_path, language="en")
-        transcribed_text = clean_transcription(result["text"])
-        
-        return jsonify({"text": transcribed_text})
-    except Exception as e:
-        return jsonify({"error": f"Speech recognition error: {str(e)}"}), 500
-
-# Use Waitress for Production Server
-if __name__ == "__main__":
-    serve(app, host="0.0.0.0", port=7860)