This package locally downloads the metadata for all video in a user's Youtube history and performs a quick analysis. The results are then displayed in the browser.
The final results of the analysis should look pretty similar to mine:
https://jessime.github.io/youtube_gh_pages/
This script uses Python 3.x. If you don't have Python, I would recommend downloading it from Anaconda.
Copy or clone this package from Github.
Open the Terminal/Command Line and navigate to where you copied the package:
$ cd path/to/copied/directory
Then, just run:
$ pip install -r requirements.txt
to install the dependencies.
There are currently two options for downloading data.
The prefered method is to use Google Takeout.
Make sure YouTube and YouTube Music is checked and follow directions to download the zip file.
It'll take 10-30 minutes to recieve an email from Google saying your job is done.
Unzip your downloaded file and pass it as a command line parameter:
$ python youtube_history.py --takeout /path/to/Takeout
The specific file we're looking for is YouTube and YouTube Music/history/watch-history.html.
So make sure at least that one file exists in the Takeout directory.
If you're more of a gambler, you can try using youtube-dl to download your data.
Just do:
$ python youtube_history.py
and you'll be prompted for your Google username and password if you haven't downloaded the raw metadata yet.
These are used only by youtube-dl and not saved in any way (even locally).
This is a "gamble" because youtube-dl is constantly fighting Google bot detection and losing more often than not.
So your login attempts may trigger 2-Factor Authentication or fail completely.
To help avoid triggering Google's protective measures, you can rate-limit your requests.
This will slow down download time, but increase the likelihood of being able to download everything.
To delay requests for a second, do:
$ python youtube_history.py -d 1
To specify any non-default directory for the data, run:
$ python youtube_history.py -o /path/to/empty/data/directory/
Once raw metadata is downloaded to the default data directory, this step is skipped in future runs to save time.
If you want to download a second dataset, you'll have to point the script to a new directory.
(Or you can manually clear out the default /data folder).
Feel free to direct any questions or comments to the Issues page of the repository.
- Add a pip package and command line tool.
- Add more complicated analyses, like Reddit's best method.