ask for the youtube api from the user

app.py

@@ -10,24 +10,11 @@ from transformers import AutoModelForAudioClassification, AutoFeatureExtractor
 from transformers import AutoTokenizer, AutoModelForSequenceClassification
 from transformers import BlipProcessor, BlipForConditionalGeneration
 import cv2
-import browser_cookie3
-
-def get_youtube_cookies(browser):
-    if browser == 'Chrome':
-        return browser_cookie3.chrome()
-    elif browser == 'Firefox':
-        return browser_cookie3.firefox()
-    elif browser == 'Edge':
-        return browser_cookie3.edge()
-    elif browser == 'Brave':
-        return browser_cookie3.brave()
-    else:
-        raise ValueError("Unsupported browser")
-
-def download_youtube_video(video_url, browser):
-    cookies = get_youtube_cookies(browser)
+
+YOUTUBE_API_KEY = os.getenv('YOUTUBE_API_KEY')
+
+def download_youtube_video(video_url, api_key):
     ydl_opts = {
-        'cookiefile': cookies,
         'format': 'bestvideo+bestaudio',
         'outtmpl': os.path.join('./', '%(title)s.%(ext)s'),
     }
@@ -60,7 +47,7 @@ def convert_mp3_to_wav(mp3_path):
 def process_text(text):
     model_name = "cardiffnlp/twitter-roberta-base-emotion"
     emotion_labels = ['anger', 'joy', 'optimism', 'sad']
-    
+
     tokenizer = AutoTokenizer.from_pretrained(model_name)
     model = AutoModelForSequenceClassification.from_pretrained(model_name)
 
@@ -68,12 +55,12 @@ def process_text(text):
     with torch.no_grad():
         outputs = model(**inputs)
         logits = outputs.logits
-    
+
     emotion_probs = torch.softmax(logits, dim=-1).squeeze()
     predicted_emotion = emotion_labels[torch.argmax(emotion_probs)]
-    
+
     emotion_dict = {emotion_labels[i]: emotion_probs[i].item() for i in range(len(emotion_labels))}
-    
+
     return emotion_dict, predicted_emotion
 
 def preprocess_frame(frame):
@@ -89,11 +76,11 @@ def generate_caption(pixel_values):
 def predict_emotions(caption):
     inputs = emotion_tokenizer(caption, return_tensors='pt', truncation=True, padding=True)
     outputs = emotion_model(**inputs)
-    
+
     emotion_probs = torch.softmax(outputs.logits, dim=1)
-    
+
     predicted_emotions = {label: prob.item() for label, prob in zip(emotion_labels, emotion_probs[0])}
-    
+
     return predicted_emotions
 
 caption_model_name = "Salesforce/blip-image-captioning-base"
@@ -104,22 +91,17 @@ emotion_model_name = "j-hartmann/emotion-english-distilroberta-base"
 emotion_tokenizer = AutoTokenizer.from_pretrained(emotion_model_name)
 emotion_model = AutoModelForSequenceClassification.from_pretrained(emotion_model_name)
 
-def analyze_video(video_url, browser):
+def analyze_video(video_url):
     global output_path
     output_path = './'
-    
-    video_path = download_youtube_video(video_url, browser)
 
+    video_path = download_youtube_video(video_url, YOUTUBE_API_KEY)
     mp4_path = convert_to_mp4(video_path)
-
     audio_path = extract_audio_from_video(mp4_path)
-
     audio_wav_path = convert_mp3_to_wav(audio_path)
 
     model_whisper = whisper.load_model("base")
-
     result_whisper = model_whisper.transcribe(audio_wav_path)
-
     transcript = result_whisper['text']
 
     emotion_dict_text, predicted_emotion_text = process_text(transcript)
@@ -128,9 +110,7 @@ def analyze_video(video_url, browser):
     emotion_vectors_video = []
 
     video_capture = cv2.VideoCapture(mp4_path)
-
     total_frames_video = int(video_capture.get(cv2.CAP_PROP_FRAME_COUNT))
-
     frame_count_video = 0
 
     while video_capture.isOpened():
@@ -150,21 +130,28 @@ def analyze_video(video_url, browser):
     video_capture.release()
 
     average_emotion_vector_video = np.mean(emotion_vectors_video, axis=0)
-
     combined_emotion_vector_final = np.concatenate((np.array(list(emotion_dict_text.values())), average_emotion_vector_video))
-
     final_most_predicted_index = np.argmax(combined_emotion_vector_final)
-
     final_most_predicted_emotion = list(emotion_dict_text.keys())[final_most_predicted_index]
 
     return transcript, predicted_emotion_text, final_most_predicted_emotion
 
-iface = gr.Interface(fn=analyze_video,
-                    inputs=[gr.Textbox(label="YouTube Video URL"), 
-                            gr.Dropdown(label="Select Browser", choices=["Chrome", "Firefox", "Edge", "Brave"])],
-                    outputs=["text", "text", "text"],
-                    title="Multimodal Emotion Recognition",
-                    description="Enter a YouTube Video URL and select your browser to analyze emotions from both audio and visual content.")
+with gr.Blocks() as iface:
+    gr.Markdown("# 🎥 Multimodal Emotion Recognition\nUpload or enter a YouTube Video URL and analyze emotions from both audio and video frames.")
+    
+    with gr.Row():
+        video_url = gr.Textbox(label="YouTube Video URL", placeholder="Enter video URL here...", interactive=True)
+        api_key = gr.Textbox(label="YouTube API Key", placeholder="Enter your API key", type="password", interactive=True)
+    
+    with gr.Row():
+        submit_button = gr.Button("Analyze Video")
+    
+    with gr.Row():
+        transcript_output = gr.Textbox(label="Transcript", interactive=False)
+        audio_emotion_output = gr.Textbox(label="Emotion from Audio", interactive=False)
+        visual_emotion_output = gr.Textbox(label="Emotion from Video", interactive=False)
+    
+    submit_button.click(analyze_video, inputs=[video_url, api_key], outputs=[transcript_output, audio_emotion_output, visual_emotion_output])
 
 if __name__ == "__main__":
-    iface.launch()
+    iface.launch()