byigitt / transcriptor

create transcripts with youtube links on google colab using whisper ai

Transcriptor

This python project lets you create multiple transcripts with youtube links on Google Colab with Whisper AI.

Questions

Why Google Colab?

I used Google Colab because of my internet speed & free GPU usage. It works flawlessly in Turkish language (which I used this to get the transcripts on Google Oyun ve Uygulama Akademisi education videos).

Features

• Downloads multiple links on the youtube_urls.txt file.
• Creates transcripts for every mp3 file it downloaded.
• Automatically deletes mp3 files after creating its transcript.
• Uses youtube-dl as in nightly mode to remove some bugs from the new latest version.

Dependencies

• Python 3.x (which Google Colab has)
• whisper
• torch

Executing the Program

• Firstly, create yourself a google colab and change runtime type to make it as a GPU.
• After that, create a youtube_urls.txt - You can find the example in our repo.
• After creating youtube_urls.txt, use !git clone https://github.com/byigitt/transcriptor.git to get the source files.
• You need to install our dependencies main.sh - In order to get it working, you need to do !chmod 755 main.sh and do !./main.sh in order to install everything.
• After installing everything, program will open itself and do its job, you can download the .txt files and you are good to go!
• Do not forget to open the tab while it does it job, otherwise your files in colab will be deleted!

For Issues and Questions

Dont feel shy to ask your questions/problems in issues tab! You can also contribute the code in Pull requests tab.