Update app.py

app.py

@@ -14,15 +14,27 @@ def play_text(text):
     os.system(f"start {temp_file.name}")  # Windows
     return "✅ Text is being read out. Please listen and read it yourself."
 
-# Function to transcribe user's audio
+# Function to transcribe user's audio and compare with the original text
 def transcribe_audio(audio, original_text):
     recognizer = sr.Recognizer()
     with sr.AudioFile(audio) as source:
         audio_data = recognizer.record(source)
+    
     try:
         start_time = time.time()
-        # Using Google Speech Recognition (supports Hindi)
-        transcription = recognizer.recognize_google(audio_data, language="hi-IN")
+        # Split the audio into chunks (1-minute chunks in this example)
+        audio_length = len(audio_data.frame_data)
+        chunk_size = 60000  # 1 minute (60,000 ms)
+        
+        # Splitting audio data into chunks
+        chunks = [audio_data.frame_data[i:i+chunk_size] for i in range(0, audio_length, chunk_size)]
+        
+        transcription = ""
+        for chunk in chunks:
+            audio_chunk = sr.AudioData(chunk, audio_data.sample_rate, audio_data.sample_width)
+            # Using Google Speech Recognition (supports Hindi)
+            transcription += recognizer.recognize_google(audio_chunk, language="hi-IN") + " "
+        
         end_time = time.time()
         
         # Calculate Accuracy
@@ -33,14 +45,19 @@ def transcribe_audio(audio, original_text):
         
         # Calculate speed
         duration = end_time - start_time  # time to process (not speaking time)
-        # Better: estimate speaking time from audio length if needed (advanced)
-        
         speed = round(len(transcribed_words) / duration, 2)  # words per second
         
+        # Compare words and highlight mistakes
+        wrong_words = []
+        for i, word in enumerate(original_words):
+            if i >= len(transcribed_words) or word != transcribed_words[i]:
+                wrong_words.append(f"🔴 {word}")
+        
         result = {
             "📝 Transcribed Text": transcription,
             "🎯 Accuracy (%)": accuracy,
-            "⏱️ Speaking Speed (words/sec)": speed
+            "⏱️ Speaking Speed (words/sec)": speed,
+            "❌ Incorrect Words": ' '.join(wrong_words) if wrong_words else "None"
         }
         return result
     except Exception as e:
@@ -67,5 +84,3 @@ with gr.Blocks() as app:
 
 # Launch the app
 app.launch()
-
-