TheTVDB - TV and Film Metadata Database

TheTVDB - TV and Film Metadata Database is the Broadcasting catalog's descriptive-layer record: a fan-contributed, community-curated catalogue operated by Whip Media that describes itself as the most accurate source for TV and film, and that pushes everything contributors add out to other sites, mobile apps and devices. Its surface covers series, seasons, episodes, artwork, movies, people, companies, awards, genres, languages, countries, content ratings and lists, plus a tagging layer that classifies how records relate. This is the metadata underneath media centres - Plex and Kodi are built on it.

What makes the catalogue distinctive is curation depth at the edges of television: aliases and translated names for regional broadcasts, absolute-number ordering for shows watched in production order, finale markers so end-of-run features need no heuristics, and a change feed that reports merges when duplicate records collapse.

Get a sample of this dataset to see series and episode records laid against your use case.

Two series records and one episode, shaped exactly as they deliver:

# SeriesBaseRecord
id              : 121361
name            : The Thick of It
slug            : the-thick-of-it
country         : GB          originalLanguage: eng
status          : Ended
firstAired      : 2005-05-19   lastAired: 2012-10-20
averageRuntime  : 29           score    : 4.2131

# SeriesBaseRecord
id              : 73141
name            : Mystery Science Theater 3000
country         : US          status: Ended
firstAired      : 1988-11-24   lastAired: 1999-08-08
averageRuntime  : 91

# EpisodeBaseRecord (seriesId 121361)
aired           : 2009-11-21   runtime: 29
seasonNumber    : 3            number : 1
absoluteNumber  : 17           finaleType: (null)

Read the anatomy rather than the individual titles. A British political satire with a 29-minute average runtime and a clean Ended status; an American show whose 91-minute average runtime immediately separates it from standard half-hour comedies; and an episode placed three ways at once - season 3, episode 1, absolute 17 - which is precisely the redundancy that lets ordering logic survive re-edits and special episodes. seriesId joins the episode back to its parent without a lookup table.

Sixteen documented fields carry every delivered row - eight describing a series, five placing an episode inside it, three carrying curation state. A further group (overview translations, artwork references, ordering flags, merge targets) is defined across the same records but was not value-sampled during the cataloging pass, so those fold under additional fields on request.

• Geography: Global. country and originalLanguage sit on every series record, and name and overview translations span dozens of language codes, so a Korean drama's French release resolves to the same series as its Seoul broadcast.
• Temporal: The full historical archive of curated television and film, plus forward nextAired values where schedules are known, and a change feed recording edits and merges over time.
• Granularity: Series, season, episode, artwork, person, company, award and list level - one stable identifier per entity, so artwork sets and cast lists hang off the same key as the title.

That combination - historical depth, forward dates and entity-level grain - is what makes this the reference catalogue for episode guides rather than a schedule grid or a ratings panel. For actual audience measurement the neighbours are BARB Weekly Top 50 Shows and Viewing Data in the UK and the Internet Archive TV News Archive for caption-searchable news history.

API, files, or your warehouse. Daily, weekly, or hourly.

Every delivery ships the complete field dictionary above, the sample rows and the coverage profile mapped to your scope - series only, one franchise, or the whole catalogue. Because records carry a last-edited timestamp and merges surface through the change feed, teams typically take the historical backfill once and keep new and edited records rotating in thereafter, with no silent drift between syncs.

1. Media-centre and product developers build guides and discovery features on stable identifiers: series ID joins artwork, cast and episode lists; translations localize the interface without re-sourcing titles.
2. Content-acquisition and programming analysts profile franchises - season counts, runtimes, finale structure, country and language mix - before licensing conversations, using curation rather than box-office proxies.
3. Market researchers studying global content flows read country-of-origin and translation coverage as a footprint of international distribution, show by show.
4. Journalists and academics cite structured facts about programmes - first and last air dates, episode counts, ordering - instead of scraping fan wikis of unknown provenance.

Across all four, the constant is that description travels with identification: one id resolves title, aliases, translations, artwork and episode tree together.

Developers and data-product builders get the canonical join layer for anything that displays television - the reason media-centre platforms standardized on it. Data scientists and ML engineers get labeled, translated text and artwork references for recommendation and content-matching models. Market researchers and consultants get programme-level descriptors that make international catalogues comparable. Competitive intelligence and product teams track rivals' catalogue shapes by genre, country and runtime mix. Journalists, academics and students cite community-audited facts with stable identifiers behind them. Persona workflows live at data scientists x broadcasting, developers & data-product builders x broadcasting and journalists, academics & students x broadcasting.

Within TV metadata, the trade-offs run on curation model and ordering fidelity. TVmaze is the realtime schedule-first option: unauthenticated JSON, full future listings, and a share-alike licence - the head-to-head lives on our comparison page. TMDB's TV section brings daily bulk exports and deep artwork but restricts free keys to non-commercial use. TheTVDB's edge is ordering and localization discipline - absolute numbers, airs-before flags and the deepest alias/translation habit in the slice - which is why media centres lean on it.

The usual pattern is to stack them: TVmaze for schedules, TMDB for bulk artwork pulls, TheTVDB as the identity spine that keeps both consistent. Where these rank for the industry sits on our broadcasting datasets ranking.

Four notes worth having in hand:

• Community-curated, professionally consumed - Contributors add and correct the records; media-centre platforms distribute them. Curation depth is the product, which is why aliases and translations arrive populated rather than null.
• Three ways to number an episode - Season number, absolute number and airs-before/after flags coexist on one episode record, so anime seasons, specials and split cour order correctly without a second table.
• Merges leave a forwarding address - When duplicate records collapse, the change feed carries a pointer to the surviving record - incremental syncs keep their joins intact instead of silently dropping history.
• Cross-identifier lookup - Records resolve against external identifiers used by other platforms, so TheTVDB IDs map onto IDs your systems already hold.

The catalogue is community-curated, so coverage concentrates where contributors concentrate: mainstream English-language television is exhaustive, while some regional or very recent programming carries thinner translation sets until someone adds them. Exact catalogue size is not published, which is why the sample - cut to the series, genres or countries you name - settles scope questions faster than any spec sheet. Where this set stops, neighbours pick up: TMDB for daily bulk exports, TVmaze for realtime schedules, and BARB for who actually watched.