Queue Up reads two public AniList lists, works out which anime both people want to watch, and ranks those by how well each title fits both tastes. This page says where every number comes from, how the ranking works, what the static pages are built from, and what the tool cannot know.
A title enters the catalog when the dataset has an AniList id for it and a known type (TV, film, OVA, ONA or special); adult content is excluded outright, and AniList's own adult flag hides such entries from the app unless you switch them on under View. The app can rank any of the 20,641 catalog titles; this site keeps a page for 3,000 of them. The rule is written down here because it decides what search engines see. First the quality gate: a TV series, film, ONA or OVA that is finished or airing, first aired in 1990 or later, carries at least three tags and scores a median of 6.5 or better across the listing sites (6,775 titles pass). Then demand: the dataset has no viewer counts, but its tag list is the union of every listing site's tags and grows with the number of people cataloguing a title, so the gate titles are ranked by ln(1 + tag count) × median score and the top 3,000 get a page. A title keeps its page once it has one, so an address that worked keeps working. Lists exist for the 79 themes with at least 40 titles, plus lists by length and status; every title page is linked from at least one list.
Tonight builds a pool in four tiers, in order of weight: titles both of you have on Planning; titles one of you completed and scored 7/10 or better (or is rewatching) while the other has on Planning ("show them"); titles one of you has on Planning that the other has never listed ("new to one of you"); and titles you are both watching or paused at within three episodes of each other ("finish it together", picking up at the next episode). Anything either of you dropped, or completed with a score under 5/10, is out. Each candidate's score is its tier weight × fit × quality. Fit is the geometric mean of each person's taste match: a taste vector per person is built from the genres and rank-weighted tags of everything they have completed, weighted by their score (unscored titles count as 7/10) and by how rare each genre or tag is across the pair's titles (Action sits on half of everything and says little about taste; Iyashikei says a lot), then normalised; the match is the cosine between that vector and the title's own genres and tags, weighted the same way. Quality is (community score ÷ 100)² × min(1, log₁₀(popularity) ÷ 5), never below 0.15 × the score part, so a little-known title is not ruled out. Ties go to whichever entry either of you touched more recently. The roll draws with odds proportional to that score; the bracket seeds by it and pairs first-round opponents by genre, then by format and length; Rewatch lists what you both completed and both scored 8/10 or better, by combined score; Find New takes catalog titles neither of you has listed, scores their fit against your combined taste and their quality from the catalog score, and names the completed titles you both share that explain each pick; Compare is the cosine between your two taste vectors and a few counts.
The verdict on every title page is produced by rules, not by anyone's opinion, and each page names the rule that fired: a film is one sitting; 100 episodes or more is a commitment; psychological, tragedy, thriller or horror tags without comedy make it a quiet night; episodic, slice-of-life, or comedy/action runs of 26 episodes or fewer are easy to drop in; 12 episodes or fewer is a weekend; anything else is a couple of weeks. Hours are episodes × episode length from the dataset; "evenings" assume two episodes a night.
Private AniList lists are invisible. A title only one of you has ever heard of is scored on the other person's taste alone, and taste vectors are only as good as your completed list — a new account with five completed titles gets vague fits. The dataset's tags are a merge of several sites' vocabularies, so a tag can be missing or odd; the catalog is refreshed when the site is rebuilt, not live, and its date is printed on every page. Streaming links are AniList's and may be out of date or region-specific. If a title looks wrong, tell us and it will be fixed in the next build.
The core catalog the app loads is one JSON file, /catalog.json (the top 3,000 titles by score and recency: title, type, episodes, length, status, season, year, score, tags, studio, cover, MyAnimeList id). It is a derivative of anime-offline-database and is available under the same ODbL terms with attribution. A machine-readable summary of the site lives at /llms.txt.
Queue Up is a program from Somebody Decide, a small software studio operated by RapidShift, LLC. Small free programs that settle the arguments friends have. Every program follows the same four rules: public data only, no accounts, nothing about you stored on a server, and the whole method written down on a page like this.
Corrections, questions and takedown requests: a public contact address is being set up and will appear here; this page is regenerated on every build. See also Legal & privacy.
Queue Up is a free tool: enter two AniList usernames and it ranks the anime you both want to watch, rolls one for tonight, runs a bracket to settle it and finds new anime that fit both of you. List data from AniList; title data: anime-offline-database (ODbL), dataset of 2026-07-04; page updated 2026-09-23. Not affiliated with AniList or MyAnimeList. Queue Up is a Somebody Decide program. How it works · Legal & privacy · All lists