abidlabs's picture
abidlabs HF staff
changes
fe8d4db
raw
history blame
2.61 kB
import gradio as gr
from pytube import YouTube
import os
from pydub import AudioSegment
import ffmpeg
def convert_to_embed_url(youtube_url):
if "youtube.com/watch?v=" in youtube_url:
video_id = youtube_url.split("v=")[1].split("&")[0]
elif "youtu.be/" in youtube_url:
video_id = youtube_url.split("youtu.be/")[1].split("?")[0]
else:
return ""
embed_url = f"https://www.youtube.com/embed/{video_id}"
return f'<iframe width="100%" height="200" src="{embed_url}" style="border-radius:10px"></iframe>'
def download_audio_from_youtube(video_url):
try:
yt = YouTube(video_url)
audio_stream = yt.streams.filter(only_audio=True).first()
downloaded_file = audio_stream.download(".")
base, _ = os.path.splitext(downloaded_file)
mp3_file = base + '.mp3'
AudioSegment.from_file(downloaded_file).export(mp3_file, format='mp3')
os.remove(downloaded_file)
return mp3_file
except Exception as e:
gr.Error(f"An error occurred: {e}")
def convert_video_to_audio(input_file):
output_file = "audio.mp3"
try:
(
ffmpeg
.input(input_file)
.output(output_file)
.run()
)
return output_file
except ffmpeg.Error as e:
gr.Error(f"An error occurred: {e}")
def transcribe_button(source):
if source == "audio":
return gr.Button("Transcribe audio πŸ“œ")
else:
return gr.Button("Transcribe video πŸ“œ")
def generate_audio(source, source_file):
if source == "audio":
audio_file = source_file
elif source == "video":
gr.Info("Converting video to audio...")
audio_file = convert_video_to_audio(source_file)
else:
gr.Info("Downloading audio from YouTube...")
audio_file = download_audio_from_youtube(source_file)
return gr.DownloadButton(value=audio_file, interactive=True)
def generate_prompt(cleanup):
prompt = "The following is a raw transcript from an automatic transcription system. "
if not cleanup:
return gr.Textbox(visible=False)
elif "Remove typos" in cleanup:
prompt += "Fix the minor typos (e.g. misspellings, homophones) in the transcript so that the transcript reads more logically. "
if "Separate into paragraphs" in cleanup:
prompt += "Separate the transcript into paragraphs to make it more readable. "
prompt += "Don't add any extra words in your response, like 'Here is the corrected transcript:' just return the final transcript."
return gr.Textbox(visible=True, value=prompt)