CoubDownloader is a simple script to download videos (called coubs) from Coub.
Linux users might be interested in CoubDownloader's successor Gyre.
Update 2021-01-02: Gyre can now also be used on Windows, but requires an MSYS2 setup (see Gyre on Windows).
CoubDownloader is a simple download script for coub.com
Usage: coub.py [OPTIONS] INPUT [INPUT]...
Input:
URL download coub(s) from the given URL
-i, --id ID download a single coub
-l, --list PATH read coub links from a text file
-c, --channel NAME download coubs from a channel
-t, --tag TAG download coubs with the given tag
-e, --search TERM download search results for the given term
-m, --community NAME download coubs from a community
NAME as seen in the URL (e.g. animals-pets)
--hot download coubs from the hot section (default sorting)
--random download random coubs
--input-help show full input help
Input options do NOT support full URLs.
Both URLs and input options support sorting (see --input-help).
Common options:
-h, --help show this help
-q, --quiet suppress all non-error/prompt messages
-y, --yes answer all prompts with yes
-n, --no answer all prompts with no
-s, --short disable video looping
-p, --path PATH set output destination (def: '.')
-k, --keep keep the individual video/audio parts
-r, --repeat N repeat video N times (def: until audio ends)
-d, --duration TIME specify max. coub duration (FFmpeg syntax)
Download options:
--connections N max. number of connections (def: 25)
--retries N number of retries when connection is lost (def: 5)
0 to disable, <0 to retry indefinitely
--limit-num LIMIT limit max. number of downloaded coubs
Format selection:
--bestvideo download best available video quality (def)
--worstvideo download worst available video quality
--max-video FORMAT set limit for the best video format (def: higher)
Supported values: med, high, higher
--min-video FORMAT set limit for the worst video format (def: med)
Supported values: med, high, higher
--bestaudio download best available audio quality (def)
--worstaudio download worst available audio quality
--aac prefer AAC over higher quality MP3 audio
--aac-strict only download AAC audio (never MP3)
--share download 'share' video (shorter and includes audio)
Channel options:
--recoubs include recoubs during channel downloads (def)
--no-recoubs exclude recoubs during channel downloads
--only-recoubs only download recoubs during channel downloads
Preview options:
--preview COMMAND play finished coub via the given command
--no-preview explicitly disable coub preview
Misc. options:
--audio-only only download audio streams
--video-only only download video streams
--write-list FILE write all parsed coub links to FILE
--use-archive FILE use FILE to keep track of already downloaded coubs
Output:
--ext EXTENSION merge output with the given extension (def: mkv)
ignored if no merge is required
-o, --output FORMAT save output with the given template (def: %id%)
Special strings:
%id% - coub ID (identifier in the URL)
%title% - coub title
%creation% - creation date/time
%community% - coub community
%channel% - channel title
%tags% - all tags (separated by _)
Other strings will be interpreted literally.
This option has no influence on the file extension.
Starting with v3.7 standalone Windows builds are provided on the release page. These executables don't require an existing Python environment and the GUI version is a standalone application. FFmpeg is still a mandatory requirement.
Because I lack the necessary build environment.
Because I lack the resources to target a variety of distros individually and my attempts to create universal Linux builds were unsuccessful.
The main problem is the usage of external tools like FFmpeg (but also media players with the --preview option) via subprocess.run. They either spam the terminal output with warnings or outright fail because of library version mismatches between my build environment and the user's system. It would've required me to ship Linux builds with static FFmpeg binaries and either remove the --preview option or make it the user's responsibility to get it working. And even then there's still a plethora of problems regarding the GUI version, which I couldn't reliably get to work across a handful of mainstream distros (Debian, Ubuntu, CentOS, Linux Mint, etc.).
- Python >= 3.7
- FFmpeg
Standalone builds only require FFmpeg.
- aiohttp for asynchronous execution (strongly recommended)
- colorama for colorized terminal output on Windows (you should also install it if you want to use
coub-gui.py) - Gooey to run
coub-gui.py(be sure to install it with wxPython < 4.1.0)
CoubDownloader's defaults can be changed via a configuration file.
Simply create a text file with the name coub.conf in the same location as the script and specify custom defaults in the following format:
OPTION = value
Lines starting with # get treated as comments.
For all available options and their allowed values, please take a look at the example configuration file.
Accessible via coub.py --input-help
CoubDownloader Full Input Help
Contents
========
1. Input Types
2. Input Methods
3. Sorting
1. Input Types
==============
-) Direct coub links
-) Lists
-) Channels
-) Searches
-) Tags
-) Communities (incl. Featured & Coub of the Day)
-) Hot section
-) Random
2. Input Methods
================
1) Direct URLs from coub.com (or list paths)
Single Coub: https://coub.com/view/1234567
List: path/to/list.txt
Channel: https://coub.com/example-channel
Search: https://coub.com/search?q=example-term
Tag: https://coub.com/tags/example-tag
Community: https://coub.com/community/example-community
Hot section: https://coub.com or https://coub.com/hot
Random: https://coub.com/random
URLs which indicate special sort orders are also supported.
2) Input option + channel name/tag/search term/etc.
Single Coub: -i 1234567 or --id 1234567
List: -l path/to/list.txt or --list path/to/list.txt
Channel: -c example-channel or --channel example-channel
Search: -e example-term or --search example-term
Tag: -t example-tag or --tag example-tag
Community: -m example-community or --community example-community
Hot section: --hot
Random: --random
3) Prefix + channel name/tag/search term/etc.
A subform of 1). Utilizes the script's ability to autocomplete/format
incomplete URLs.
Single Coub: view/1234567
Channel: example-channel
Search: search?q=example-term
Tag: tags/example-tag
Community: community/example-community
Hot section: hot
Random: random
3. Sorting
==========
Input types which return lists of coub links (e.g. channels or tags)
support custom sorting/selection methods (I will refer to both as sort
orders from now on). This is mainly useful when used in combination with
--limit-num (e.g. download the 100 most popular coubs with a given tag),
but sometimes it also changes the list of returned links drastically
(e.g. a community's most popular coubs of a month vs. a week).
Sort orders can either be specified by providing an URL that already
indicates special sorting
https://coub.com/search/likes?q=example-term
https://coub.com/tags/example-tag/views
https://coub.com/rising
or by adding it manually to the input with '#' as separator
https://coub.com/search?q=example-term#top
tags/example-tag#views_count
hot#rising
This is supported by all input methods, except the --hot option.
Please note that a manually specified sort order will overwrite the
sort order as indicated by the URL.
Supported sort orders
---------------------
Channels: most_recent (default)
most_liked
most_viewed
oldest
random
Searches: relevance (default)
top
views_count
most_recent
Tags: popular (default)
top
views_count
fresh
Communities: hot_daily
hot_weekly
hot_monthly (default)
hot_quarterly
hot_six_months
rising
fresh
top
views_count
random
Featured: recent (default)
(community) top_of_the_month
undervalued
Coub of the Day: recent (default)
(community) top
views_count
Hot section: hot_daily
hot_weekly
hot_monthly (default)
hot_quarterly
hot_six_months
rising
fresh
Random: popular (default)
top
The following points provide more in-depth information about the different input types, which didn't make it into the overview because of self-imposed space restrictions.
A link to a single coub (e.g. https://coub.com/view/123456). This is the most basic form of input and what the other input types boil down to once parsed.
Lists are files on your computer, which will be scanned for direct coub links. In order to detect a direct link, it must separated from the surrounding content via one of the following delimiters:
- whitespace
- tab
- newline
Additionally a direct link must start with 'https://coub.com/view/'.
CoubDownloader itself is also able to create lists with --write-list. When used, all parsed links will be outputted to a user-defined file. This helps to avoid additional parsing time, if a long download is split into several sessions, but also to weed out duplicate links in an already existing list.
Channels allow to download all (re)coubs from a single user. The channel options:
--recoubs--no-recoubs--only-recoubs
allow fine-grained control over what type of coub to download. As they are global options, they apply to all channels for a single script instance. Story download is not supported.
Searches provide the same results as on Coub's website, with the exception that it is not possible to search for channels.
Using general search terms can potentially return tens of thousands of coub links. The usage of --limit-num is encouraged.
Provides a random collection of coubs. A request returns max. 1000 coubs, but several requests can be made during one script invocation (e.g. using the --random option thrice).
The now following input types only return a limited number of links. This is indirectly enforced by the API, although I can't say if it is done on purpose or a bug (pages >99 redirect to page 1).
Each of the following input types provides at most 2475 direct links.
Tags (just like searches) are the same as the user will know them from Coub's website.
Please note that the default sort order (by popularity) provides less results than all the other sort orders.
There are currently 19 supported communities.
The following list shows the names of the supported communities (as seen on Coub's website). In parenthesis are the internally used names and what should be used as input for the script.
- Animals & Pets (animals-pets)
- Mashup (mashup)
- Anime (anime)
- Movies & TV (movies)
- Gaming (gaming)
- Cartoons (cartoons)
- Art & Design (art)
- Music (music)
- News & Politics (news)
- Sports (sports)
- Science & Technology (science-technology)
- Celebrity (celebrity)
- Nature & Travel (nature-travel)
- Fashion & Beauty (fashion)
- Dance (dance)
- Auto & Technique (cars)
- NSFW (nsfw)
- Featured (featured)
- Coub of the Day (coub-of-the-day)
The default sort order for most communities (most popular coubs of the month) may provide less results than other sort orders.
Featured and Coub of the Day have unique sort orders.
The currently most popular/trending coubs.
Both
refer to the hot section and can be used as input.
The default sort order (most popular coubs of the month) may provide less results than other sort orders.
A basic GUI, powered by Gooey, is provided via coub-gui.py.
Please note that this is only a quickly thrown together frontend for convenience. The main focus of this project is the CLI tool.
It provides the same functionality as the main CLI tool, with a few notable exceptions:
- No quiet mode
- No overwrite prompt (default prompt answer is set to "no")
- No option equivalent to
--random#top(direct URL input must be used) - The output path defaults to "coubs" in the user's home directory instead of the current one
Another important difference is that coub-gui.py is NOT a standalone script. It depends on coub.py being in the same location.
Resolution is not a synonym for quality. That is what everybody with a bit of knowledge in video encoding will tell you. It is important to remember, when we discuss the quality of available video streams.
Coub usually offers 3 video streams:
- higher (~1600px width)
- high (~1280px width)
- med (~640px width)
This is also the order in which coub.py ranks those video streams. Said ranking can be limited with --min-/--max-video and the final choice depends on --best-/--worstvideo. But the question is, does this really reflect the quality of those streams? And the answer is: No
The main problem is that Coub does something very peculiar. In order to provide several available streams, they don't just scale input videos down, but also up. This is a recent change. Before the introduction of the watermarks, you would find coubs, which were only available in 360p, simply because a low resolution input video was used. Now you can find the same video in 900p as well, because they (quite noticeably) upscaled the low resolution input.
This makes it very difficult to discern which stream actually offers the best quality. Not because of the used compression settings (Coub reencodes look equally bad at all resolutions), but because of the details lost in the scaling process. Usually you want to get the stream, which has the resolution closest to the source video and scale it yourself during playback. A properly configured media player can produce far better results than Coub's blurry upscaling. But even if blurry upscaling is preferred (e.g. to mitigate compression artifacts), you will save a considerate amount of disk space by letting your media player handle the scaling process.
But how to apply that knowledge in practice? How do you know which stream to download? Well, I'm still looking for an answer myself. The only accurate method is to download all versions and compare them yourself. Details are a good indicator to look out for. Pick the stream which still has the most of them left. Alternatively you can also compare, which stream looks the sharpest to you.
Unfortunately there's no clear rule, which would help to automate the decision process. However based on my personal experience I can give you the following hint. The best quality is often 'high', sometimes 'med' and almost never 'higher'.
I'd like to quickly address how coub.py handles AAC audio, because it might be a bit confusing.
The script gathers potential audio streams and chooses the final link based on --best-/worstaudio.
The audio streams are by default ranked in the following order:
- MP3@160Kbps VBR
- AAC@128Kbps CBR
- MP3@128Kbps CBR
So it's pretty uncommon to get AAC audio. Only if the high quality MP3 audio stream isn't present, will the script switch to AAC. However, some users might prefer AAC audio, which is why I added the --aac and --aac-strict options.
With --aac the ranking will simply change to:
- AAC@128Kbps CBR
- MP3@160Kbps VBR
- MP3@128Kbps CBR
It basically tells the script to rank AAC higher than anything else. On the other hand --aac-strict reduces the ranking to:
- AAC@128Kbps CBR
Either an AAC stream is present or the audio will be entirely missing. This ensures AAC audio under any circumstances. Another way to look at it is that --aac tries to download AAC with MP3 as fallback, while --aac-strict gets rid of the fallback.
To make matters even more complicated, some users might not want AAC audio at all. This is hopefully only a small demographic (after all AAC support is thorough and it does compress a lot better than MP3), but the script is still able to cater to this group. There's no extra command line option, but look for the following lines inside the script and change AAC to 0.
# How much to prefer AAC audio
# 0 -> never download AAC audio
# 1 -> rank it between low and high quality MP3
# 2 -> prefer AAC, use MP3 fallback
# 3 -> either AAC or no audio
AAC = 1
Now AAC audio will be completely ignored and the script only serves MP3 audio (like the old version).
Another special new option is --share. Coub now offers a video version primarily targeted at people, who want to share coubs. These videos already contain both video (~1280px width, sometimes ~360px) and audio (AAC@128Kbps CBR) and don't require further muxing. Videos downloaded with --share come as MP4.
WARNING: There's a danger of overwriting share videos with video-only streams, as they both come as MP4.
The downside is that the audio is often considerably shorter than the other available audio streams. Sometimes that's beneficial, as Coub tends to loop the audio for short tracks (e.g. 2x 25 sec. audio). Often you just lose a lot of audio duration though (e.g. 3 min. song gets reduced to 20 sec.).
Also because of the special property of the share version, there are some pitfalls to look out for. Many options become useless, when used together with --share
--short--keep--repeat--duration- all other format selection options
--audio-onlyand--video-only(throws error)
There's no fallback for share videos. If the share version is not yet available, then the script will count the coub as unavailable.
- no error handling or retries if server disconnects during input parsing
- losing your internet connection while downloading will stall the script indefinitely
- some enabling switches (e.g. --keep) don't have a disabling counterpart, so activating their options via a config means you can't disable them at runtime
- using a CLI media player to preview audio files may lead to its keyboard shortcuts not working
- Python 3.8 and newer will print tracebacks (only warnings) when the user interrupts the program (see the Wiki for more infos)
- (GUI only) progress messages don't use monospace fonts on Windows
- (GUI only) having wrong values defined in your config will prevent the GUI from opening (open it from the command line to see the error)
- (GUI only) Gooey seems to be broken for WxPython 4.1.0