IMDb Non-Commercial Datasets

It is the reference catalog of screen entertainment, delivered as seven coordinated tables instead of one sprawling dump. Title basics carries every movie, short, TV series, episode, video and video game IMDb indexes - twelve million-plus titles - with its tconst, type, year, runtime and up to three genres. Around it sit ratings, alternate titles, crew, principals and episodes, plus a people table covering sixteen million-plus individuals.

The design decision worth admiring: everything joins on two stable identifiers, tconst for titles and nconst for people. A rating finds its title, a credit finds its person, an episode finds its series, and none of it requires fuzzy name matching. Datadory normalizes the whole set into query-ready form, so "what is this title and who made it" becomes a column lookup - see how movies & entertainment data families stack.

One row per title in basics, with the rating arriving on the shared identifier:

# one row per title in title.basics; ratings, localized
# titles, crew, principals and episodes join on tconst
file            : title.basics
tconst          : tt0111161
titleType       : movie
primaryTitle    : The Shawshank Redemption
startYear       : 1994
runtimeMinutes  : 142
genres          : Drama

# the same tconst carries its crowd score:
file            : title.ratings
averageRating   : 9.3
numVotes        : 2900000

# people arrive on their own nconst; episodes tie back
# to a parent series with season and episode numbers;
# a backslash-N marks a missing value everywhere

The pattern worth noticing: each table holds one grain and refuses to mix them, so the catalog stays flat while the depth lives in the joins. Rows reflect the August 2026 review pass; a fresh pull arrives current to the day it ships.

Thirty fields span the seven tables, verified against the published dictionary during the research pass. The title spine: tconst, titleType, primaryTitle, originalTitle, isAdult, startYear, endYear, runtimeMinutes and genres. Scoring adds averageRating - a weighted average of individual user ratings - and numVotes. Alternate titles contribute region, language, types, ordering and isOriginalTitle, so one work's life under other flags becomes filterable.

Credits and people complete it: directors and writers as nconst arrays, then nconst, category, job, characters and ordering per principal credit. Episodes tie installments to their parents with parentTconst, seasonNumber and episodeNumber, and name basics rounds out with primaryName, birthYear, deathYear, primaryProfession and knownForTitles. The full dictionary follows in the table below.

Geography - global by construction: every film, television series, short, video and video game production IMDb indexes worldwide, whatever country it came out of.

Temporal - from cinema's beginnings to currently releasing titles, with end years marking closed runs. Snapshots land complete whenever they ship, and the cadence they reach you on is yours to set - which turns a moving catalog into a panel the moment you accumulate pulls. Pair the slice with GroupLens MovieLens Datasets for interaction history or Box Office Mojo for grosses when a question needs behavior or money next to the record.

Granularity - title-level and person-level records, with credit-level principals underneath and episode-level linkage tying installments to their series. Three grains, three tables, no ambiguity.

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

• Recommendation and personalization systems - weighted ratings and vote counts on stable identifiers remain the canonical cold-start signal for recommender models; see ml model training.
• Catalog enrichment - year, runtime, genres, cast and localized titles attach to any corpus you already own via a one-hop join.
• Localization research - alternate titles by region, language and type show how a single work travels under different names market to market.
• Talent and credit analysis - principals with categories and roles, plus professions and known-for titles per person, reconstruct careers credit by credit.
• Demand proxies - vote volume complements box office and streaming signals for a fuller interest picture; see demand forecasting.