app.py

import cv2
import gradio as gr
import google.generativeai as genai
import os
import PIL.Image

# Configure the API key for Google Generative AI
genai.configure(api_key=os.environ.get("GOOGLE_API_KEY"))

# Define the Generative AI model
model = genai.GenerativeModel('gemini-1.5-flash')

# Function to capture frames from a video
def frame_capture(video_path, num_frames=5):
    vidObj = cv2.VideoCapture(video_path)
    frames = []
    total_frames = int(vidObj.get(cv2.CAP_PROP_FRAME_COUNT))
    frame_step = max(1, total_frames // num_frames)
    count = 0
    
    while len(frames) < num_frames:
        vidObj.set(cv2.CAP_PROP_POS_FRAMES, count)
        success, image = vidObj.read()
        if not success:
            break
        frames.append(image)
        count += frame_step
    
    vidObj.release()
    return frames

# Function to generate text descriptions for frames
def generate_descriptions_for_frames(video_path):
    frames = frame_capture(video_path)
    images = [PIL.Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)) for frame in frames]
    
    prompt = "Describe what is happening in each of these frames."
    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

# Helper function to format descriptions
def format_descriptions(descriptions):
    return ' '.join(descriptions).strip()

# Define Gradio interface
video_input = gr.Video(label="Upload Video", autoplay=True)
output_text = gr.Textbox(label="What's in this video")

# Create Gradio app
gr.Interface(fn=generate_descriptions_for_frames, inputs=video_input, outputs=output_text, title="Video Content Detection System").launch()