ytscribe

A simple library that allows you to transcribe any single person YouTube video into a text file using OpenAI's Whisper module.

Setup

1. Clone this library and cd into the ytscribe folder:

git clone https://github.com/vamsiuppala/ytscribe.git
cd ytscribe

2. Install dependencies

pip install -r requirements.txt

From PIP

You can also install the library from pip

pip install ytscribe

Transcribe YouTube video

Step 1. Get the link to the YouTube video you're interested in

1. This package has one main module that takes YouTube link and simply transcribes it.
2. Pick a video that only has a single speaker since Whisper doesn't differentiate between multiple speakers
3. As an example, let's pick one of MKBHD's videos from AutoFocus - https://www.youtube.com/watch?v=CRhjL9X2yKA

Step 2. Pass on the YouTube video link

video_url = "https://www.youtube.com/watch?v=CRhjL9X2yKA"
directory = "files"
from ytscribe import transcribe

# transcribe
transcribe.download_video_and_transcribe(video_url, directory, output_filename = 'title', whisper_model = 'base')

The above bit should create a folder called files and add the transcribed YouTube video as a txt file.

Speed

On my 16GB M1 Pro Macbook Pro, a 20 minute video took around 2 minutes to get downloaded and transcribed

Certificate issues

If you run into ssl certification troubles, this command helped me

import ssl
ssl._create_default_https_context = ssl._create_unverified_context