|
import whisper |
|
import yt_dlp |
|
import gradio as gr |
|
import os |
|
import re |
|
|
|
model = whisper.load_model("base") |
|
|
|
def get_audio(url): |
|
try: |
|
ydl_opts = { |
|
'format': 'bestaudio/best', |
|
'noplaylist': True, |
|
'quiet': True, |
|
'outtmpl': '%(title)s.%(ext)s' |
|
} |
|
with yt_dlp.YoutubeDL(ydl_opts) as ydl: |
|
info = ydl.extract_info(url, download=True) |
|
|
|
audio_file = ydl.prepare_filename(info) |
|
return audio_file |
|
except Exception as e: |
|
raise gr.Error(f"Exception: {e}") |
|
|
|
def get_text(url): |
|
try: |
|
if url != '': |
|
audio_file = get_audio(url) |
|
result = model.transcribe(audio_file) |
|
return result['text'].strip() |
|
else: |
|
return "Please enter a YouTube video URL." |
|
except Exception as e: |
|
raise gr.Error(f"Exception: {e}") |
|
|
|
def get_summary(article): |
|
try: |
|
first_sentences = ' '.join(re.split(r'(?<=[.:;])\s', article)[:5]) |
|
return first_sentences |
|
except Exception as e: |
|
raise gr.Error(f"Exception: {e}") |
|
|
|
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.</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>Transcription takes 5-10 seconds per minute of the video (bad audio/hard accents slow it down a bit). #patience<br />If you have time while waiting, check out my <a href=https://www.artificial-intelligence.blog target=_blank>AI blog</a> (opens in new tab).</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_transcribe.click(get_text, inputs=input_text_url, outputs=output_text_transcribe) |
|
|
|
demo.queue(default_enabled=True).launch(debug=True) |
