fiftyone.utils.youtube#

Utilities for working with YouTube <https://youtube.com>.

Copyright 2017-2025, Voxel51, Inc.
voxel51.com

Functions:

download_youtube_videos(urls[, ...])

Downloads a list of YouTube videos.
fiftyone.utils.youtube.download_youtube_videos(urls, download_dir=None, video_paths=None, clip_segments=None, ext='.mp4', only_progressive=True, resolution='highest', max_videos=None, num_workers=None, skip_failures=True, quiet=False)#

Downloads a list of YouTube videos.

The urls argument accepts a list of YouTube “watch” URLs:

urls = [
    "https://www.youtube.com/watch?v=-0URMJE8_PB",
    ...,
]

Use either the download_dir or video_paths argument to specify where to download each video.

You can use the optional clip_segments argument to specify a specific segment, in seconds, of each video to download:

clip_segments = [
    (10, 25),
    (11.1, 20.2),
    None,               # entire video
    (None, 8.0),        # through beginning of video
    (8.0, None),        # through end of video
    ...
]

You can also use the optional ext and resolution arguments to specify a desired video codec and resolution to download, if possible.

YouTube videos are regularly taken down. Therefore, this method provides an optional max_videos argument that you can use in conjunction with skip_failures=True and a large list of possibly non-existent videos in urls in cases where you need a certain number of videos to be successfully downloaded but are willing to tolerate failures.

Parameters:

  • urls – a list of YouTube URLs to download

  • download_dir (None) – a directory in which to store the downloaded videos

  • video_paths (None) – a list of paths to which to download the videos. When downloading entire videos, a stream matching the video format implied by each file’s extension is downloaded, if available, or else the extension of the video path is changed to match the available stream’s format

  • clip_segments (None) – a list of (first, last) tuples defining a specific segment of each video to download

  • ext (".mp4") – a video format to download for each video, if possible. Only applicable when a download_dir is used. This format will be respected if such a stream exists, otherwise the format of the best available stream is used. Set this value to None if you want to download the stream with the best match for resolution and progressive regardless of format

  • only_progressive (True) – whether to only download progressive streams, if possible. Progressive streams contain both audio and video tracks and are typically only available at <= 720p resolution

  • resolution ("highest") –

    a desired stream resolution to download. This filter is applied after respecting the desired video format and only_progressive restriction, if applicable. The supported values are:

    • "highest" (default): download the highest resolution stream

    • "lowest": download the lowest resolution stream

    • A target resolution like "1080p". In this case, the stream whose resolution is closest to this target value is downloaded

  • max_videos (None) – the maximum number of videos to successfully download. By default, all videos are downloaded

  • num_workers (None) – a suggested number of threads/processes to use when downloading videos

  • skip_failures (True) – whether to gracefully continue without raising an error if a video cannot be downloaded

  • quiet (False) – whether to suppress logging, except for a progress bar

Returns:

a tuple of

  • downloaded: a dict mapping integer indexes into urls to paths of successfully downloaded videos

  • errors: a dict mapping integer indexes into urls to error messages for videos that were attempted to be downloaded, but failed

IN THIS ARTICLE