app.py

import streamlit as st
from pytube import YouTube
import yt_dlp
import os


# from faster_whisper import WhisperModel
# model = WhisperModel("large-v2", device="cpu", device_index=0,  compute_type="float16")

DESCRIPTION = """
Welcome to the **YouTube Video summary** powered by Llama-2 models.
"""
st.title("YouTube Video summary")
st.markdown(DESCRIPTION)

def get_video_title(youtube_url: str) -> str:
    yt = YouTube(youtube_url)
    embed_url = f"https://www.youtube.com/embed/{yt.video_id}"
    embed_html = f'<iframe src="{embed_url}" frameborder="0" allowfullscreen></iframe>'
    return yt.title, embed_html


def initialize_session_state():
    if "youtube_url" not in st.session_state:
        st.session_state.youtube_url = ""
    if "model_choice" not in st.session_state:
        st.session_state.model_choice = "Llama2-70b"
    if "setup_done" not in st.session_state:
        st.session_state.setup_done = False
    if "doneYoutubeurl" not in st.session_state:
        st.session_state.doneYoutubeurl = ""

def sidebar():
    with st.sidebar:
        st.markdown("Enter the YouTube Video URL below🔗")
        st.session_state.youtube_url = st.text_input("YouTube Video URL:")

        if st.session_state.youtube_url:
            # Get the video title
            video_title, embed_html = get_video_title(st.session_state.youtube_url)
            st.markdown(f"### {video_title}")

            # Embed the video
            st.markdown(embed_html, unsafe_allow_html=True)

sidebar()
initialize_session_state()

ydl_opts = {
    'outtmpl': 'demo.m4a',
    'format': 'm4a/bestaudio/best',
    # ℹ️ See help(yt_dlp.postprocessor) for a list of available Postprocessors and their arguments
    'postprocessors': [{  # Extract audio using ffmpeg
        'key': 'FFmpegExtractAudio',
        'preferredcodec': 'm4a',
    }],
}

if st.session_state.youtube_url:
    with st.status("Get video Audio..."):
        if os.path.exists('demo.m4a'):
            os.remove('demo.m4a')
        with yt_dlp.YoutubeDL(ydl_opts) as ydl:
            error_code = ydl.download([st.session_state.youtube_url])

    audio_file = open('demo.m4a', 'rb')
    audio_bytes = audio_file.read()
    st.audio(audio_bytes, format='audio/ogg')

    
    # segments, info = model.transcribe("demo.m4a", beam_size=5)
    # st.markdown("Detected language '%s' with probability %f" % (info.language, info.language_probability))

    # full_response = ""
    # message_placeholder = st.empty()
    # for segment in segments:
    #     # print("[%.2fs -> %.2fs] %s" % (segment.start, segment.end, segment.text))
    #     full_response += segment.text + " "
    #     st.write(segment.text)
        # message_placeholder.markdown(full_response + "▌")
    # message_placeholder.markdown(full_response)