RV

Huh?

RV is a video downloader with a lot of features, most of which are filters for fine-tuning your search

How to use

Python 3.10 or greater required

• RV is a cmdline tool, no GUI
• See requirements.txt for additional dependencies. Install with:
  • python -m pip install -r requirements.txt

Install as a module

• cd rv
• python -m pip install .
• python -m rv [options...] OR simply
• rv [options...]

Without installing

• cd rv

• Run either:

  • python rv/__main__.py [options...] OR
  • rv.cmd [options...] (Windows)
  • rv.sh [options...] (Linux)

• For bug reports, questions and feature requests use our issue tracker

Search

• RV provides advanced searching functionality (rv pages). Search is performed using extended website native API
• There are 4 search arguments available each corresponding to different parts of search API. It's possible to utilize all four at once but usually a single one is enough
  • -search <STRING> - search using raw string, matching all words (see below). Concatenate using -:
    • -search after-hours
  • -search_tag <TAGS>, -search_cat <CATEGORIES>, -search_art <ARTISTS> - search using one or more tag/category/artist names (see below). Concatenate using ,:
    • -search_tag 1girl,side_view
    • -search_cat cyberpunk_2077
    • -search_art hydrafxx

Search rules

• In addition to ability to use more than one tag (or category/artist) name at once there is also a combining logic - match all or any of them, for example -search_tag 1female,1girl by default will try to match posts having both tags, to make it match any of those tags you must specify tags search rule argument
  • -search_rule_tag any
• Same with categories and artists
  • -search_cat cyberpunk_2077,overwatch -search_rule_cat any
  • -search_art hydrafxx,medeister -search_rule_art any
• Note that overall search always obeys AND rule:
  • search string AND ANY_OF/ALL the tags AND ANY_OF/ALL the artists AND ANY_OF/ALL the categories
• How raw string search really works: before trying to match word(s) in the search string server performs alias expansion which may lead to some unexpected results, aliases list is unavailable but some of them are straight counterintuitive - too many post may get matched for seemingly no reason. Words in provided string may match title, tags, categories, author, uploader or description, pretty much anything. Matching is partial so word 'all' will match 'tall', 'calling' and everything else containing this symbol sequence

Filters

Native filters

• In addition to multiple search parameters it is also possible to filter out unwanted results natively. Syntax:
  • -blacklist [type1:]name1[,[type2:]name2,...], where <typeN> is either a (artist), c (category) or t (tag), and <nameN> is (non-)wildcarded value string
  • <nameN> must be a valid value of type <typeN> (see below for possible values)
  • Each type can be used multiple times
  • If type is omitted then name is matched against all possible types and must match at least one of them
  • Examples:
  • Filter out tag 2d
    • -blacklist t:2d
  • Filter out all tags starting with fur or scal and all categories starting with rob
    • -blacklist t:fur*,t:scal*,c:rob*
  • Filter out any tags, categories and artists with name containing zombie:
    • -blacklist *zombie*

Post-search filters

• Initial search results / ids list can be then filtered further using extra tags (see below)

Tags. Categories. Artists. Extra tags

• rv_tags.list file contains all existing tags for current version, same with rv_cats.list (categories) and rv_arts.list (artists) files. Any tag/category/artist you use is required to be valid and every extra tag needs to be a valid tag, category or artist. That is, unless you also utilize...
• Wildcards. In any extra tag you can use symbols ? and * for any symbol and any number of any symbols repectively
• If even more advanced approach is required you can also use regular expressions. To prevent syntax conflicts following regex symbols must be escaped using `: ()?*.,-+ as well as (?: sequence (`(?:). Example: *[1`-5]`+`(finger{1`,3}|girl`)s`?`.`* converts to regex ^.*[1-5]+(?:finger{1,3}|girl)s?.*$. Notes:
  • No need to specify group as non-capturing
  • Some characters don't need escaping, like | or [ there
  • You can combine wildcards and regular expressions within the same extra tag. Note how first * is converted as wildcard symbol while the ending `.`* specified explicitly as regex converts to the same characters pair
  • ` character is used for escaping because it isn't contained in any tag, artist or category name
• Fuzzy extra tags (containing wildcards or regular expressions) aren't validated, they can be anything
• Using extra tags is the only way to filter by uploader (user) name. There are two options available:
  • Special syntax: -u:<NAME>. The -u: (or positive u:) prefix explicitly tells to use this extra tag as uploader filter only, <NAME> here is your normal extra tag so it can also contain wildcards/regex
  • Cmd option --check-uploader which extends all extra tags application to also affect uploader in addition to tags/categories/artists, use this with caution
• What makes extra tags different from tags/categories/artists is these tags or -tags are being used as filters instead of search params, normal tags/categories/artists are passed using their own search argument (see full help) and all unknown arguments are automatically considered extra tags
• All spaces must_be_replaced_with_underscores ‒ all tag / category / artist names are unified this way for convenience

Additional info

1. OR / AND groups:

• OR group is a parenthesized tilda (~) -separated group of tags
  • (<tag1>~<tag2>~...~<tagN>)
  • video matching any of the tags in OR group is considered matching that group
• AND group is a parenthesized comma (,) -separated group of tags. Only negative AND group is possible ‒ to filter out videos having this unwanted tags combination
  • -(<tag1>,<tag2>,...,<tagN>)
  • video matching all tags in AND group is considered matching that group

2. --download-scenario explained in detail:

• Syntax: --download-scenario SCRIPT / -script SCRIPT
• Scenario (script) is used to separate videos matching different sets of tags into different folders in a single pass
• SCRIPT is a semicolon-separated sequence of '<subfolder>: <args...>' groups (subqueries)
• SCRIPT always contains spaces hence has to be escaped by quotes:
  • rv ids <args>... -script "sub1: tags1; sub2: tags2 ..."
• Typically each next subquery is better exclude all required tags from previous one and retain excluded tags, so you know exactly what file goes where. But excluding previous required tags is optional ‒ first matching subquery is used and if some item didn't match previous sub there is no point checking those tags again. Subquery order matters. Also, -tags contained in every subquery can be safely moved outside of script
  • ... -script "s1: a b (c~d) -e; s2: -a -b -c -d -e f g (h~i); s3: -a -b -c -d -e -f -g -h -i k" << full script
  • ... -script "s1: a b (c~d) -e; s2: f g (h~i) -e; s3: k -e" << no redundant excludes
  • ... -script "s1: a b (c~d); s2: f g (h~i); s3: k" -e << "-e" moved outside of script
• Besides tags each subquery can also have -quality set ‒ videos matching that subquery will be downloaded in this quality
• Subquery can also have --use-id-sequence flag set (see below) and match video ids
• Subquery can also have its own -duration filter - only videos having duration within -duration bounds will be downloaded to that subfolder
• You can also set --untagged-policy always for one subquery

3. Downloading a set of videos

• There are two ways to download a set of ids instead of id range:
• Using id sequence:
  • Syntax: --use-id-sequence / -seq, rv ids only (or download scenario subquery)
  • The sequence itself is an extra tag in a form of OR group of ids:
    • (id=<id1>~id=<id2>~...~id=<idN>)
  • Id sequence is used instead of id range, you can't use both
    • rv ids <args>... -seq (id=1337~id=9999~id=1001)
• By passing video links directly:
  • Syntax: --use-link-sequence / -links, rv ids only
  • Links are extracted from extra tags turned into ids
  • Id sequence is used instead of id range/sequence, you can't use both
    • rv ids <args>... -links https://... https://...

4. File naming

• File names are generated based on video title and tags:
• Base template: <prefix>_<id>_(<score>)_<title>_(<tags>).<ext>. It can be adjusted via -naming argument
• Non-descriptive or way-too-long tags will be dropped
• If resulting file full path is too long to fit into 240 symbols, first the tags will be gradually dropped; if not enough title will be shrunk to fit; general advice is to not download to folders way too deep down the folder tree

5. Using 'file' mode

• File mode becomes useful when your cmdline string becomes really long. For example: Windows string buffer for console input is about 32767 characters long but standard cmd.exe buffer can only fit about 8192 characters, powershell ‒ about 16384. File mode can be used with shorter cmdline string as well
• File mode is activated by using 'file' subcommand and has a single option which is -path to a text file containing actual cmdline arguments:
  • rv file -path <FILEPATH>
• Target file has to be structured as follows:
  • all arguments and values must be separated: one argument or value per line
  • quotes you would normally use in console window to escape argument value must be removed
  • only actual arguments needed

    pages
    -start
    1
    -end
    20
    -path
    H:/long/folder name/with spaces (no quotes)/
    --log-level
    trace
    -script
    s1: (script~is~a~single~value); s2: -no_quotes_here_either
    

6. Unfinished files policy

• Unexpected fatal errors, Ctrl-C and other mishaps will cause download(s) to end abruptly
• By default when app manages to exit gracefully all unfinished files get deleted, at the same time all existing files are automatically considered completed
• To check and resume existing unfinished files use --continue-mode (or -continue) option. This may be slower for non-empty folders due to additional network requests but safer in case of complex queries
• To keep unfinished files use --keep-unfinished (or -unfinish) option. It acts as --continue-mode helper so it's recommended to use either both or none at all

7. Interrupt & resume

• When downloading at large sometimes resulting download queue is so big it's impossible to process within reasonable time period and the process will be inevitably interrupted
• There is only one way to make a 'safe' interruption which is to tap 'q' twice. This will signal the scanner to skip the rest of its queue and after all remaining downloads get processed the app will exit normally
• To be able to resume without running the whole search process again use --store-continue-cmdfile option. Once initial video queue is formed a special 'continue' file will be stored and periodically updated in base download destination folder
• Continue file contains cmdline arguments required to resume download, all provided parameters / options / download scenario / extra tags are preserved
• It is strongly recommended to also include --continue-mode and --keep-unfinished options when using continue file
• If download actually finishes without interruption stored continue file is automatically deleted
• Continue file has to be used with rv file (see using 'file' mode above)

8. Wildcards in search

• Once familiar enough with existing tags/categories/artists lists one may want to go advanced and use wildcards in typed search (not search string)
• Syntax is the same, search rules apply as normal, but tag/category/artist name containing wildcard symbol has to match at least one existing tag/category/artist name in list
• Example: -search_tag ?girl*,?boy* will be automatically expanded to:
  • -search_tag 1girl,2girls,3girls,...,1boy1girl,2boys,3boys,3boys1girl,...,6girls

9. Scanning for unpublished posts

• There is always a number of posts which haven't become public yet, many of them don't pass a review and never become available
• Unpublished posts cannot be searched for so using direct link is the only option
• rv ids can automatically scan past maximum available post ID until it reaches the actual maximum post ID available.
• Syntax: -lookahead <AMOUNT>. <AMOUNT> here is the number of sequential empty post IDs to assume the end of existing posts. Example:
  • rv ids -start 3500000 -count 500 ... -lookahead 100 - scan until 3,500,500 and continue indefinetely until 100 post requests in a row return 'not found' error

Examples

1. Pages

• All videos by a single tag:
  • rv pages -pages 9999 -search_tag TAG1
• Up to 48 videos with both tags present from a single author in 1080p, save to a custom location:
  • rv pages -pages 2 -path PATH -quality 1080p -search_art ARTIST -search_tag TAG1,TAG2
• Up to 24 videos on page 3 with any of 3 tags from any of 2 authors under any of 2 categories, exclude any kind of vore or fart, in best quality, with minimum score of 100 and minimum rating of 90%, use proxy, save to a custom location, save tags, log everything, use shortest names for files:
  • rv pages -log trace -start 3 -pages 1 -path PATH -proxy https://127.0.0.1:222 -tdump -quality 2160p -minscore 100 -minrating 90 -search_cat CATEGORY1,CAT_EGORY2 -search_art ART_IST1,ARTIST2 -search_tag TAG1,TAG2,TAG3 -search_rule_cat any -search_rule_art any -search_rule_tag any -naming 0 -*vore -fart*
• All videos uploaded by a user, if tagged with either of 2 desired tags, in best quality, sorted into subfolders by several desired (known) authors, putting remaining videos into a separate folder, setup for interrupt & continue:
  • rv pages -pages 9999 -path PATH --store-continue-cmdfile -quality 2160p -uploader USER_ID (TAG1~TAG2) -script "name1: AUTHOR1; name2: AUTHOR2; name3: AUTHOR3; rest: * -utp always"
• All videos from artist's page, if tagged with a desired tag, in best quality:
  • rv pages -pages 999 -path PATH -quality 2160p -model ARTIST_NAME TAG1

2. Ids

• All existing videos in range:
  • rv ids -start 3200000 -count 100
  • rv ids -start 3200000 -end 3200099
• You can use the majority of arguments from pages examples. The only argument that is unique to rv ids is --use-id-sequence (-seq), see above where it's explained in detail

Common mistakes

There are several ways you can select videos to download and some of them, despite being seemingly logical, are terribly ineffective:

1. Using rv ids to process greater id range

• rv ids <args...> -start 1 -end 3500000
  • approximate time such query takes is 40+ days
  • videos 236 - 3,045,049 don't exist

Solution 1: split your query into smaller ones:

• rv ids <args...> -start 1 -end 235
• rv ids <args...> -start 3045050 -end 3100000
• etc.

Solution 2: use rv pages (listed videos):

• rv pages <args...> -start 1 -pages 9999

2. Using extra tags to download only certain tag / category / artist / uploader

• rv pages <args...> -start 1 -pages 9999 ARTIST
• rv ids <args...> -start ... -end ... (TAG1~TAG2)
• etc.
  • you may end up having to check every video

Solution 1: use native search functionality instead (search result pages):

• rv pages <args...> -start 1 -pages 99 -search_art ARTIST
• rv pages <args...> -start 1 -pages 99 -search_tag TAG1,TAG2 -search_rule_tag any

Solution 2: to search by uploader use -uploader argument (user video pages):

• rv pages <args...> -start 1 -pages 99 -uploader USER_ID
  • USER_ID ‒ integer, can be found in member's profile page address