Update app.py

app.py

@@ -127,7 +127,6 @@ pipe_asr = pipeline(
     return_timestamps=True
 )
 
-# Function to handle audio input, transcribe, fetch from Neo4j, and generate audio response
 # Function to handle audio input, transcribe, fetch from Neo4j, and generate audio response
 def transcribe_and_respond(audio):
     if audio is None:
@@ -144,11 +143,8 @@ def transcribe_and_respond(audio):
         y = torchaudio.functional.resample(torch.tensor(y), orig_freq=sr, new_freq=target_sr).numpy()
         sr = target_sr
 
-    # Prepare input_features for Whisper model
-    input_features = processor(y, sampling_rate=sr, return_tensors="pt").input_features
-
     # Transcribe the audio using Whisper with English language setting
-    result = pipe_asr({"input_features": input_features, "language": "en"}, return_timestamps=False)
+    result = pipe_asr({"raw": y, "sampling_rate": sr}, return_timestamps=False)
     question = result.get("text", "")
 
     # Log the transcribed text for debugging
@@ -169,6 +165,7 @@ def transcribe_and_respond(audio):
     
     return audio_path, response_text
 
+
 # Function to clear the transcription state
 def clear_transcription_state():
     return None, None