Update app.py

app.py

@@ -4,26 +4,30 @@ import torch
 import speech_recognition as sr
 from transformers import pipeline
 from gtts import gTTS
-import time
+import re
 
 app = Flask(__name__)
 recognizer = sr.Recognizer()
 
-# Load Hugging Face Whisper Model for Speech-to-Text
+# Load Whisper Model (English Only)
 device = "cuda" if torch.cuda.is_available() else "cpu"
 speech_to_text = pipeline("automatic-speech-recognition", model="openai/whisper-base", device=0 if device == "cuda" else -1)
 
-# Function to generate and save voice prompts
+# Function to generate and save voice prompts in English
 def generate_audio(text, filename):
     tts = gTTS(text=text, lang="en")
     tts.save(filename)
 
-# Generate all prompts before starting
+# Generate all voice prompts before starting
 generate_audio("Welcome to Biryani Hub.", "static/welcome.mp3")
 generate_audio("Tell me your name.", "static/ask_name.mp3")
 generate_audio("Please provide your email.", "static/ask_email.mp3")
 generate_audio("Thank you for registration.", "static/thank_you.mp3")
 
+# Function to clean text and remove non-English characters
+def clean_text(text):
+    return re.sub(r'[^a-zA-Z0-9@.\s]', '', text)  # Allow only English letters, numbers, @, and spaces
+
 @app.route("/")
 def home():
     return render_template("index.html")
@@ -38,8 +42,10 @@ def process_audio():
     audio_file.save(audio_path)
 
     try:
-        text = speech_to_text(audio_path)["text"]
-        return jsonify({"text": text})
+        # Force Whisper to transcribe in English only
+        text = speech_to_text(audio_path, generate_kwargs={"language": "en"})["text"]
+        cleaned_text = clean_text(text)  # Clean non-English characters
+        return jsonify({"text": cleaned_text})
     except Exception as e:
         return jsonify({"error": str(e)}), 500