YCDL creates an SQLite3 database of Youtube channels and their videos, and serves it out of a web server.
The database acts as a permanent archive of video metadata including title, description, duration, view count, and more. Even if a video or channel is deleted from Youtube, you will still have this information. Perfect for keeping track of unlisted videos, too.
The thumbnails, however, are not stored in the database, but you can use utilities\download_thumbnails.py
to download them.
Note: At this time, refreshing a channel in YCDL will update video titles, descriptions, and view counts with their current values. If you refresh a channel after they have changed their video's title or description you will lose the previous value.
When I discover a channel, I like to watch through the videos over the course of weeks or months. Within Youtube's own interface, there is no good way to filter videos you've watched from videos you haven't. Scrolling through all of a channel's videos to find ones you haven't seen is tough.
In YCDL, videos start off as pending and you can mark them as ignore or download. The pending page always acts as your "to-watch" list.
On Youtube's subscription page, there is a button to hide a video from the list. I would often press this hide button after watching a video, only to find it come back a few days later. Don't get me started on live broadcasts or premieres -- hiding those was never reliable. YCDL makes watching my subscriptions much easier.
YCDL does not perform the downloading of videos itself. youtube-dl is the tool for that. When you click on the download button, it will create an empty file called xxxxxxxxxxx.ytqueue
in the directory specified by the ycdl.json
config file. You should create a separate shell / Python script that watches for ytqueue files and calls youtube-dl with your preferred arguments.
The reason for this is that youtube-dl is extremely configurable. Every user might prefer a completely different set of arguments and formatting. Rather than attempting to provide an interface for that in YCDL, my goal is to get you the video IDs so you can pass them into your favorite youtube-dl configuration.
I wrote a program called q.py to watch my ytqueue folder and call youtube-dl with each of the video IDs in an endless loop.
- Web interface with video embeds
- Commandline interface for scripted use
- "New videos" feed page where newest videos from all channels are listed in order (wow, no algorithm!)
- Sort videos by date, duration, views, or random
- Background thread will refresh channels over time
- Mark channels as auto-download or auto-ignore
- Automatically remove shorts from feed
- Free yourself from Youtube's awful recommendation system
You are responsible for your own youtube_credentials.py
file in a folder on your PYTHONPATH
. This file must have a function get_youtube_key
. YCDL will import youtube_credentials
and call youtube_credentials.get_youtube_key()
with no arguments. It should return a Youtube API key string. Here is how to get one:
- Go to https://console.developers.google.com/.
- Create a project using the menu in the upper left.
- From the project's dashboard, click "Enable APIs and Services".
- Search for and choose the latest YouTube Data API.
- On the left bar, click "Credentials".
- Click "Create credentials" and choose "API key". In my experience they all start with "AIzaSy".
- Return this value from
get_youtube_key
.
YCDL has a core backend package and separate frontends that use it. These frontend applications will use import ycdl
to access the backend code. Therefore, the ycdl
package needs to be in the right place for Python to find it for import
.
-
Run
pip install -r requirements.txt --upgrade
. -
Make a new folder somewhere on your computer, and add this folder to your
PYTHONPATH
environment variable. For example, I might useD:\pythonpath
or~/pythonpath
. Close and re-open your Command Prompt / Terminal so it reloads the environment variables. -
Place your
youtube_credentials.py
file inside that folder. -
Run
python -c "import youtube_credentials; print(youtube_credentials)"
to confirm. If you see an ImportError or ModuleNotFoundError, double check your pythonpath. -
Add a symlink to the ycdl folder into the same folder where you placed
youtube_credentials.py
:The repository you are looking at now is
D:\Git\YCDL
or~/Git/YCDL
. You can see the folder calledycdl
.Windows:
mklink /d fakepath realpath
For examplemklink /d "D:\pythonpath\ycdl" "D:\Git\YCDL\ycdl"
Linux:
ln --symbolic realpath fakepath
For exampleln --symbolic "~/Git/YCDL" "~/pythonpath/ycdl"
-
Run
python -c "import ycdl; print(ycdl)"
to confirm.
In order to prevent the accidental creation of databases, you must first use ycdl_cli.py init
to create your database.
-
cd
to the folder where you'd like to create the YCDL database. -
Run
python frontends/ycdl_cli.py --help
to learn about the available commands. -
Run
python frontends/ycdl_cli.py init
to create a database in the current directory.
Note: Do not cd
into the frontends folder. Stay in the folder that contains your _ycdl
database and specify the full path of the frontend launcher. For example:
Windows:
D:\somewhere> python D:\Git\YCDL\frontends\ycdl_cli.py
Linux:
/somewhere $ python /Git/YCDL/frontends/ycdl_cli.py
It is expected that you create a shortcut file or launch script so you don't have to type the whole filepath every time.
-
Use
ycdl_cli init
to create the database in the desired directory. -
Run
python frontends/ycdl_flask/ycdl_flask_dev.py [port]
to launch the flask server. Port defaults to 5000 if not provided. -
Open your web browser to
localhost:<port>
.
Note: Do not cd
into the frontends folder. Stay in the folder that contains your _ycdl
database and specify the full path of the frontend launcher. For example:
Windows:
D:\somewhere> python D:\Git\YCDL\frontends\ycdl_flask\ycdl_flask_dev.py 5001
Linux:
/somewhere $ python /Git/YCDL/frontends/ycdl_flask/ycdl_flask_dev.py 5001
Add --help
to learn the arguments.
It is expected that you create a shortcut file or launch script so you don't have to type the whole filepath every time.
-
Use
ycdl_cli init
to create the database in the desired directory. -
Run
python frontends/ycdl_repl.py
to launch the Python interpreter with the YCDLDB pre-loaded into a variable calledY
. Try things likeY.get_videos
.
Note: Do not cd
into the frontends folder. Stay in the folder that contains your _ycdl
database and specify the full path of the frontend launcher. For example:
Windows:
D:\somewhere> python D:\Git\YCDL\frontends\ycdl_repl.py
Linux:
/somewhere $ python /Git/YCDL/frontends/ycdl_repl.py
It is expected that you create a shortcut file or launch script so you don't have to type the whole filepath every time.
Here is a javascript bookmarklet that you can click while on youtube.com to be redirected to the same URL on your YCDL server. Mainly for /channel
and /watch
URLs.
javascript:
document.location = document.location.href.replace('https://www.youtube.com', 'https://localhost:5000');
false;
Replace 5000 with the port on which you choose to run YCDL.
I use the following filters to prevent annoying elements from appearing inside the embedded youtube player:
! This is the thing that pops up while the video is paused
youtube.com##.ytp-scroll-min.ytp-pause-overlay
! These are the recommendations that appear after the video is over
youtube.com##.ytp-suggestion-set
- Keep permanent record of titles and descriptions.
- Progress indicator for channel refresh.
https://git.voussoir.net/voussoir/ycdl
https://github.com/voussoir/ycdl