Update app.py

app.py

@@ -29,33 +29,44 @@ def frame_capture(video_path, num_frames=5):
     vidObj.release()
     return frames
 
-# Function to generate text descriptions for frames
-def generate_descriptions_for_frames(video_path, user_prompt):
+# Function to generate text descriptions for frames or answer a specific question
+def analyze_video(video_path, user_question):
     frames = frame_capture(video_path)
     images = [PIL.Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)) for frame in frames]
     
-    prompt = f"Describe what is happening in each of these frames in this video sequentially. {user_prompt}"
+    if user_question.strip():
+        prompt = f"Based on these video frames, {user_question}"
+    else:
+        prompt = "Describe what is happening in each of these frames in this video sequentially."
+    
     images_with_prompt = [prompt] + images
     
     responses = model.generate_content(images_with_prompt)
     descriptions = [response.text for response in responses]
     
-    formatted_description = format_descriptions(descriptions)
-    return formatted_description
+    return descriptions[-1] if user_question.strip() else format_descriptions(descriptions)
 
 # Helper function to format descriptions
 def format_descriptions(descriptions):
     return ' '.join(descriptions).strip()
 
+# Function to handle chat interaction
+def chat_interaction(video_path, chatbot, user_message):
+    response = analyze_video(video_path, user_message)
+    chatbot.append((user_message, response))
+    return "", chatbot
+
 # Define Gradio interface
 video_input = gr.Video(label="Upload Video", autoplay=True)
+chatbot = gr.Chatbot(label="Video Analysis Chatbot")
 user_input = gr.Textbox(label="Ask something specific about the video", placeholder="E.g., Are there any cars in this video?")
-output_text = gr.Textbox(label="What's in this video")
 
 # Create Gradio app
-gr.Interface(
-    fn=generate_descriptions_for_frames,
-    inputs=[video_input, user_input],
-    outputs=output_text,
-    title="Interactive Video Analysis System"
-).launch()
+with gr.Blocks() as demo:
+    with gr.Column():
+        video_input.render()
+        chatbot.render()
+        user_input.render()
+        user_input.submit(fn=chat_interaction, inputs=[video_input, chatbot, user_input], outputs=[user_input, chatbot])
+
+demo.launch()