Source
Query TikTok videos and transcripts with SQL
TikTok public-catalogue videos — about 7.1 billion videos from 38.8 million creators (measured 2026-08-27) with descriptions, hashtags, engagement counts, and 1.26 billion caption-transcript rows — queryable with read-only SQL by creator and video.
Query surfaces
tiktok.videostiktok.transcripts
Public TikTok video pages and auto-caption tracks; source-native video and transcript records.
Best for
a creator’s full public catalogue by handle — the indexed access path
spoken content: caption transcripts per (video, language), creator captions and TikTok ASR distinguished
hashtag, music, and duration facets on every video row
Fields
idhandleprovenance_handleauthor_idauthor_nameauthor_verifiedcreate_timedescriptionhashtagsmusicdurationplay_countlike_countcomment_countshare_countcollect_counttranscripttranscript_sourceurlfetched_atsourcecrawler_versionsource_fileloaded_at
Read live from the schema registry for tiktok.videos — 24 columns. Every relation's full contract is served by /v1/scry/schema.
Queries
A creator’s newest videos
Handle is the indexed key; collapse pre-merge duplicates on id.
SELECT id, create_time, description, play_count, like_count, url
FROM (
SELECT * FROM tiktok.videos
WHERE handle = 'khaby.lame'
ORDER BY loaded_at DESC
LIMIT 1 BY id
)
ORDER BY create_time DESC
LIMIT 20;
What a creator says on camera
Transcripts by handle, one row per video and language, with the caption source.
SELECT video_id, lang, source, is_autogen, char_len, substring(text, 1, 160) AS excerpt
FROM (
SELECT * FROM tiktok.transcripts
WHERE handle = 'khaby.lame'
ORDER BY loaded_at DESC
LIMIT 1 BY video_id, lang
)
ORDER BY create_time DESC
LIMIT 20;