import whisper
from pytube import YouTube
from transformers import pipeline
import gradio as gr
import os
import re

model = whisper.load_model("base")
# model = pipeline(model="AlexMo/FIFA_WC22_WINNER_LANGUAGE_MODEL")
summarizer = pipeline("summarization")


def getAudio(url):
    link = YouTube(url)
    video = link.streams.filter(only_audio=True).first()
    file = video.download(output_path=".")
    base, ext = os.path.splitext(file)
    file_ext = base + '.mp3'
    os.rename(file, file_ext)
    return file_ext


def getText(url):
    if url != '':
        output_text_transcribe = ''
    res = model.transcribe(getAudio(url))
    return res['text'].strip()


def getSummary(article):
    header = ' '.join(re.split(r'(?<=[.:;])\s', article)[:5])
    b = summarizer(header, min_length=15, max_length=120, do_sample=False)
    b = b[0]['summary_text'].replace(' .', '.').strip()

    return b


with gr.Blocks() as demo:
    gr.Markdown(
        "<h1><center>Free Fast YouTube URL Video to Text using <a href=https://openai.com/blog/whisper/ target=_blank>OpenAI's Whisper</a> Model</center></h1>")
    gr.Markdown(
        "<center>Enter the link of any YouTube video to generate a text transcript of the video and then create a summary of the video transcript.</center>")
    gr.Markdown(
        "<center><b>'Whisper is a neural net that approaches human level robustness and accuracy on English speech recognition.'</b></center>")
    gr.Markdown(
        "<center>Generating the transcript takes 5-10 seconds per minute of the video</center>")

    input_text_url = gr.Textbox(placeholder='Youtube video URL', label='URL')
    result_button_transcribe = gr.Button('1. Transcribe')
    output_text_transcribe = gr.Textbox(placeholder='Transcript of the YouTube video.', label='Transcript')

    result_button_summary = gr.Button('2. Create Summary')
    output_text_summary = gr.Textbox(placeholder='Summary of the YouTube video transcript.', label='Summary')

    result_button_transcribe.click(getText, inputs=input_text_url, outputs=output_text_transcribe)
    result_button_summary.click(getSummary, inputs=output_text_transcribe, outputs=output_text_summary)

demo.launch(debug=True)