Update app.py

app.py

@@ -15,7 +15,7 @@ from collections import Counter
 import os
 
 # Load necessary models and files
-text_model = load_model('model_for_text_emotion_updated(1).keras')  # Load your text emotion model
+text_model = load_model('model_for_text_emotion_updated(1).keras')  # Load text emotion model
 with open('tokenizer.json') as json_file:
     tokenizer = tokenizer_from_json(json.load(json_file))  # Tokenizer for text emotion
 audio_model = load_model('my_model.h5')  # Load audio emotion model
@@ -44,6 +44,9 @@ def predict_text_emotion(text):
 
 # Extract audio features and predict emotion
 def extract_audio_features(audio_data, sample_rate):
+    if not isinstance(audio_data, np.ndarray):
+        audio_data = np.array(audio_data)  # Ensure it's a NumPy array
+
     mfcc = np.mean(librosa.feature.mfcc(y=audio_data, sr=sample_rate).T, axis=0)
     return np.expand_dims(mfcc, axis=0)
 
@@ -111,7 +114,12 @@ def transcribe_and_predict_video(video_path):
     image_emotion = process_video(video_path)
 
     # Predict emotion from audio (sound-based)
-    sample_rate, audio_data = librosa.load(audio_file, sr=None)
+    audio_data, sample_rate = librosa.load(audio_file, sr=None)
+
+    # Debugging print statements
+    print(f"Type of audio_data: {type(audio_data)}")  # Ensure audio_data is numpy.ndarray
+    print(f"Sample rate: {sample_rate}")
+
     audio_emotion = predict_audio_emotion(audio_data, sample_rate)
 
     # Combine the detected emotions for final output (you could average them or choose the most common)