Spaces:
Running
Running
| # Importing required libraries and modules | |
| import streamlit as st # Import streamlit | |
| from pytube import YouTube # Import YouTube from pytube | |
| from pytube import extract # Import extract from pytube | |
| from youtube_transcript_api import YouTubeTranscriptApi # Import YouTubeTranscriptApi | |
| import whisper # Import whisper library | |
| import pandas as pd # Import pandas | |
| import numpy as np # Import numpy | |
| import argostranslate.package # Import argostranslate.package | |
| import argostranslate.translate # Import argostranslate.translate | |
| import os # Import os | |
| from gtts import gTTS # Import gTTS | |
| st.title('Foreign Whispers - YouTube Video Translation') # Header of application | |
| # Take YouTube video link as input | |
| video_url = st.text_input('Paste the link of a YouTube video:', | |
| value = "https://www.youtube.com/watch?v=2ylerrtD_4Y") | |
| # Dropdown box for selecting target language | |
| language = st.selectbox('Translate video to:', | |
| ("Chinese", "French", "German", "Greek", "Hindi", "Italian", | |
| "Japanese", "Korean", "Portuguese", "Russian", "Spanish", "Turkish"), | |
| placeholder="Select a language...") | |
| translate_button = st.button("Translate Video") # Button to start translation process | |
| if translate_button: # Upon button activation | |
| st.divider() # Divider | |
| video_id = extract.video_id(video_url) # Acquire video ID to get transcript | |
| vid = YouTube(video_url) # Create YouTube object of video | |
| video = vid.streams.get_highest_resolution() # Get video information on highest offered resolution | |
| video = np.asarray(video) # Turn video in numpyarray | |
| st.subheader(vid.streams[0].title) # Display video title | |
| argostranslate.package.update_package_index() # Update Argos package index | |
| packages = argostranslate.package.get_available_packages() # Install Argos Translate package | |
| in_language = "en" # Set input language as English | |
| out_language = "es" # Default output language to Spanish | |
| if language == "Chinese": | |
| out_language = "zh" # Set output language as Chinese | |
| if language == "French": | |
| out_language = "fr" # Set output language as French | |
| if language == "German": | |
| out_language = "de" # Set output language as German | |
| if language == "Greek": | |
| out_language = "el" # Set output language as Greek | |
| if language == "Hindi": | |
| out_language = "hi" # Set output language as Hindi | |
| if language == "Italian": | |
| out_language = "it" # Set output language as Italian | |
| if language == "Japanese": | |
| out_language = "ja" # Set output language as Japanese | |
| if language == "Korean": | |
| out_language = "ko" # Set output language as Korean | |
| if language == "Portuguese": | |
| out_language = "pt" # Set output language as Portuguese | |
| if language == "Russian": | |
| out_language = "ru" # Set output language as Russian | |
| if language == "Spanish": | |
| out_language = "es" # Set output language as Spanish | |
| if language == "Turkish": | |
| out_language = "tr" # Set output language as Turkish | |
| package_to_install = next( # Initialize languages to install | |
| filter(lambda x: x.from_code == in_language # Install input language (English) | |
| and x.to_code == out_language, packages)) # Install output language | |
| argostranslate.package.install_from_path(package_to_install.download()) # Install languages | |
| transcript = YouTubeTranscriptApi.get_transcript(video_id) # Acquire video transcript | |
| timestamps = [] # Initialize timestamp list | |
| original_transcript = [] # Initialize original transcript list | |
| translated_transcript = [] # Initialize translation list | |
| for phrase in transcript: # For every phrase in the transcript | |
| timestamps.append(phrase['start']) # Add timestamp to timestamp list | |
| original_transcript.append(phrase['text']) # Add phrase to transcript list | |
| translated_text = argostranslate.translate.translate(phrase['text'], in_language, out_language) # Perform translation | |
| translated_transcript.append(translated_text) # Add translated phrase to list | |
| # Construct dataframe of timestamps, original phrases and translated phrases | |
| data = {'Timestamp': timestamps, | |
| 'Original Transcript': original_transcript, | |
| 'Translated Transcript': translated_transcript} | |
| df = pd.DataFrame(data) # Create a pandas dataframe with transcript data | |
| st.dataframe(df) # Print out table of the dataframe to the user | |
| full_translated_transcript = ' '.join(translated_transcript) # Piece phrases together into a full transcript | |
| tts = gTTS(text=full_translated_transcript, lang=out_language) # Perform TTS on full transcript | |
| translated_audio = 'translated_transcript_audio.mp3' # Set filename of TTS audio | |
| tts.save(translated_audio) # Save TTS audio to file | |
| st.audio("translated_transcript_audio.mp3") # Display audio player for TTS audio |