PyTube

Pythonic bindings for the YouTube API.

Prerequisites

Python 2.5+

Get Some Videos!

Display the 50 most recent video ids and titles from the Mahalo Baking channel

import pytube
client = pytube.Client('my-app-identifier')
videos = client.user_videos('mahalobaking')
for video in videos[:50]:
    sys.stdout.write('%s %s\n' % (video.id, video.title))

Streams are cached!

Streams try to do as few API queries as possible

vs = client.user_videos('mahalobaking')     # does not hit the YouTube API
top_ten = vs[:10]                           # actually grabs the first page of results (50) from API
first = vs[0]                               # returns instantly from local cache!

Updating video Metadata

You can edit any of the following properties of a video:

• title
• description
• category
• keywords
• access_control
• private

Just call Video.update() when you are done editing the video to push the updated metadata back into youtube.

v = authenticated_client.user_videos()[0]   # get a video this client owns
v.description = "My new description"        # change the video description
v.keywords.append('awesome')                # add a keyword
v.update()                                  # push the updated metadata back to youtube

Motivation

The gdata API obscures the data you are actually interested in; suppose that you wanted to find out who was the author of a given video; you might write a function like this:

gdata:

import gdata.youtube.service
def get_author(video_id):
    yt_service = gdata.youtube.service.YouTubeService()
    yt_service.source = 'my-example-application'
    entry = yt_service.GetYouTubeVideoEntry(video_id=video_id)
    return entry.author[0].name.text

pytube:

import pytube
def get_author(video_id)
    entry = pytube.Client('my-example-application').video(video_id)
    return entry.author

Introspecting a gdata entry isn't terribly helpful: All of the relevant data is buried in deep levels of object hierarchy. PyTube tries to expose as much data as possible in python native data structures. Try looking at a video object's __dict__!

Planned Features

• OAuth Authentication Support

• Performing multiple API queries simultaneously for increased speed when iterating across large collections

• Support for Playlist Operations

Known Issues

• WHERE ARE THE TESTS

• A video entry without a view_count may either have zero views or it may have it's statistics protected. (http://bit.ly/hWwk40)

• len(Stream) will return the total length of the stream (number of videos in a channel, number of search results, etc), but only the first 1000 of these results are iterable. This is a restriction of the YouTube API.

• video_id_from_youtube_url doesn't support channel urls (example: http://www.youtube.com/user/beyonceVEVO#p/u/0/4m1EFMoRFvY )

• Data objects contain references to their client; pickling a VideoStream will also pickle the associated client, which may inadvertently cause authentication tokens to be preserved.

• Timezone handling is hard. video publish dates are returned in UTC but insight data is for days in Pacific time.

• Updating a video from a pytube client can cause the loss of location / geodata on the video

Authors

Noah Silas (noah@mahalo.com)

Kai Powell (kai@mahalo.com)