Update app.py

app.py

@@ -76,11 +76,12 @@ def transcribe_audio(audio_file):
         # Convert to float32 numpy array
         audio_input = audio_input.astype(np.float32)
 
-        # Process in chunks of 30 seconds
+        # Process in chunks of 30 seconds with overlap
         chunk_length = 30 * sr
+        overlap = 5 * sr  # 5 seconds overlap
         transcriptions = []
 
-        for i in range(0, len(audio_input), chunk_length):
+        for i in range(0, len(audio_input), chunk_length - overlap):
             chunk = audio_input[i:i+chunk_length]
             input_features = processor(chunk, sampling_rate=16000, return_tensors="pt").input_features.to(device)
             predicted_ids = model.generate(input_features)
@@ -90,6 +91,7 @@ def transcribe_audio(audio_file):
         # Join all transcriptions
         full_transcription = " ".join(transcriptions)
 
+        print(f"Full transcription length: {len(full_transcription)} characters")
         return full_transcription
     except Exception as e:
         print(f"Error in transcribe_audio: {str(e)}")
@@ -104,6 +106,8 @@ def transcribe_video(url):
         # Convert audio bytes to AudioSegment
         audio = AudioSegment.from_file(io.BytesIO(audio_bytes))
         
+        print(f"Audio duration: {len(audio) / 1000} seconds")
+        
         # Save as WAV file
         with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_audio:
             audio.export(temp_audio.name, format="wav")
@@ -111,7 +115,7 @@ def transcribe_video(url):
 
         print("Starting audio transcription...")
         transcript = transcribe_audio(temp_audio_path)
-        print("Transcription completed successfully")
+        print(f"Transcription completed. Transcript length: {len(transcript)} characters")
         
         # Clean up the temporary file
         os.unlink(temp_audio_path)
@@ -123,6 +127,8 @@ def transcribe_video(url):
         return transcript
     except Exception as e:
         error_message = f"An error occurred: {str(e)}"
+        print(error_message)
+        return error_message
 
 def download_transcript(transcript):
     with tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.txt') as temp_file: