Update app.py

app.py

@@ -2,12 +2,25 @@ import whisper
 import gradio as gr
 import subprocess
 from autocorrect import Speller
+from transformers import pipeline
 
 # Load the Whisper model
 model = whisper.load_model("large")
 
 # Initialize autocorrect for Urdu
-spell = Speller(lang='ur')  # Set the language for Urdu
+spell = Speller(lang='ur')  # Urdu spelling correction
+
+# Load the transformer model for text correction
+corrector = pipeline("text-generation", model="dbmdz/bert-large-uncased-finetuned-urdu")
+
+def correct_urdu_text(text):
+    try:
+        # Use the transformer model to correct the Urdu text
+        corrected_text = corrector(text, max_length=512, num_return_sequences=1)[0]['generated_text']
+        return corrected_text
+    except Exception as e:
+        print(f"Error in text correction: {e}")
+        return text
 
 def transcribe_video(video_path):
     try:
@@ -25,7 +38,10 @@ def transcribe_video(video_path):
         # Correct the transcribed text using autocorrect
         corrected_text = spell(transcribed_text)
 
-        return corrected_text
+        # Apply further correction using transformer-based model
+        final_corrected_text = correct_urdu_text(corrected_text)
+
+        return final_corrected_text
 
     except FileNotFoundError:
         return "Error: ffmpeg is not installed or not found in the environment."
@@ -38,7 +54,7 @@ interface = gr.Interface(
     inputs=gr.Video(label="Upload your Urdu-speaking video"),
     outputs=gr.Textbox(label="Corrected Transcribed Text"),
     title="Urdu Video Transcription with Correction",
-    description="Upload a video file in Urdu, and this app will transcribe the speech and correct the text using Whisper and autocorrect.",
+    description="Upload a video file in Urdu, and this app will transcribe the speech, correct spelling and grammar using Whisper and Transformers.",
 )
 
 # Launch the app