app.py

import streamlit as st
import os
import cv2
import torch
from PIL import Image
from gtts import gTTS
from scenedetect import open_video, SceneManager, ContentDetector
from transformers import BlipProcessor, BlipForConditionalGeneration
from openai import OpenAI
# Load AI models
caption_processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
caption_model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base").to("cpu")
caption_model = torch.quantization.quantize_dynamic(caption_model, {torch.nn.Linear}, dtype=torch.qint8)
client = OpenAI(
  api_key="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"
)


# Streamlit App UI
st.title("🎥 AI-Powered Video Summarization")
st.markdown(
    """
    <style>
    .stApp {
        background-image: url("https://s29938.pcdn.co/wp-content/uploads/2019/02/Wallpaper-Kemra-WhiteSubwayTile-4-1100x1320.jpg.optimal.jpg");
        background-size: cover;
        background-position: center;
        background-repeat: no-repeat;
    }
    </style>
    """,
    unsafe_allow_html=True
)
    

uploaded_file = st.file_uploader("📤 Upload a Video File", type=["mp4"])

if uploaded_file:
    video_path = "input_video.mp4"
    with open(video_path, "wb") as f:
        f.write(uploaded_file.getbuffer())

    st.video(video_path)

    # Scene Detection & Frame Extraction
    st.write("🔍 Detecting scene changes and extracting key frames...")

    def extract_key_frames(video_path, output_folder="frames"):
        os.makedirs(output_folder, exist_ok=True)
        video = open_video(video_path)
        scene_manager = SceneManager()
        scene_manager.add_detector(ContentDetector(threshold=27.0))
        scene_manager.detect_scenes(video)
        scenes = scene_manager.get_scene_list()
        cap = cv2.VideoCapture(video_path)
        for i, (start, end) in enumerate(scenes):
            frame_time = start.get_frames()  # Extract frame at scene start
            cap.set(cv2.CAP_PROP_POS_FRAMES, frame_time)
            ret, frame = cap.read()
            
            if ret:
                frame_path = os.path.join(output_folder, f"scene_{i+1}.jpg")
                cv2.imwrite(frame_path, frame)
                print(f"Saved: {frame_path}")

        cap.release()

    extract_key_frames(video_path)

    # Caption Generation
    st.write("📝 Generating captions for extracted frames...")

    def generate_caption(image_path):
        image = Image.open(image_path).convert("RGB")
        inputs = caption_processor(image, return_tensors="pt")
        caption_ids = caption_model.generate(**inputs)
        return caption_processor.decode(caption_ids[0], skip_special_tokens=True)

    captions = []
    for filename in sorted(os.listdir("frames")):
        if filename.endswith(".jpg"):
            image_path = os.path.join("frames", filename)
            captions.append(generate_caption(image_path))

    # st.write("📄 Generated Captions:", captions)

    # Summarization
    st.write("📖 Summarizing captions using AI...")

    def summarize_captions(captions):
      prompt = f"Summarize the following sequence of video frames into a meaningful story under 500 characters:\n\n{captions}"
      
      completion = client.chat.completions.create(
          model="gpt-4o-mini",
          messages=[{"role": "system", "content": "You are an AI that summarizes video content."},
                    {"role": "user", "content": prompt}],
          max_tokens=150
      )
      return completion.choices[0].message.content
        

    summary = summarize_captions(captions)
    st.write("📌 Video Summary:", summary)

    # Text-to-Speech
    st.write("🔊 Generating voice narration...")

    def text_to_speech(text, output_audio="summary_audio.mp3"):
        tts = gTTS(text, lang="en")
        tts.save(output_audio)
    
    text_to_speech(summary)
    st.audio('summary_audio.mp3')

    # # Combine Audio & Video
    # st.write("🎬 Merging audio with the video...")

    # def add_audio_to_video(video_path, audio_path, output_video="final_video.mp4"):
    #     video = mp.VideoFileClip(video_path)
    #     audio = mp.AudioFileClip(audio_path)
    #     if audio.duration > video.duration:
    #         audio = audio.subclip(0, video.duration)
    #     final_video = video.set_audio(audio)
    #     final_video.write_videofile(output_video, codec="libx264", audio_codec="aac")

    # add_audio_to_video(video_path, "summary_audio.mp3")

    # st.video("final_video.mp4")