YouTube Transcripts

Give it YouTube video, Shorts, channel or playlist links (or plain video IDs) and it returns each transcript (subtitles, captions) as JSON: full text, timed segments, language, whether it was auto-generated, every caption language the video has, and the video's title, channel, length, publish date and views.

Sample output

One row as the API returns it. Long text is shortened here; names inside text read [name].

Download

Sample row, one field per line
Field	Value
rowType	transcript
platform	youtube
videoId	dQw4w9WgXcQ
url	https://www.youtube.com/watch?v=dQw4w9WgXcQ
title	Rick Astley - Never Gonna Give You Up (Official Video) (4K Remaster)
channelName	Rick Astley
channelId	UCuAXFkgsw1L7xaCfnd5JJOw
channelHandle	@RickAstleyYT
channelUrl	https://www.youtube.com/channel/UCuAXFkgsw1L7xaCfnd5JJOw
durationSeconds	213
publishDate	2009-10-24T23:57:33-07:00
viewCount	1,829,055,386

Call it from code or from your agent

This is the request. Small runs answer in the same call; large ones return a run ID.

TerminalOpens at launch

curl -X POST https://scrapers-api.arcadiumdata.io/v1/scrapers/youtube-transcripts/runs \
  -H "Authorization: Bearer $ARCADIUM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "input": {
      "urls": [
        "https://www.youtube.com/watch?v=dQw4w9WgXcQ"
      ],
      "captionType": "any",
      "maxVideosPerSource": 10,
      "subtitleFormat": "none"
    }
  }'

Number of transcripts

$1.5011,000 transcripts: $1.50, plus the run start fee of $0.001. A failed or empty run costs $0.

What each row holds

Every row says what it is in rowType: transcript rows are charged, coverage and error rows are free.

The transcript 31 fields

The transcript
Field	Type	What it holds
platform	string	Always 'youtube'.
videoId	string	YouTube video id, 11 characters, e.g. dQw4w9WgXcQ.
url	string	Watch page link of the video (Shorts included).
title	string	Video title.
channelName	string	Name of the channel that published the video (the publisher, not a viewer).
channelId	string	Channel id, starts with UC, e.g. UCuAXFkgsw1L7xaCfnd5JJOw.
channelHandle	string	Channel handle with the @, e.g. @RickAstleyYT. Null if YouTube did not return it.
channelUrl	string	Link to the channel page built from channelId.
durationSeconds	number	Video length in seconds.
publishDate	string	Publish date and time as YouTube gives it, ISO 8601 with offset, e.g. 2009-10-24T23:57:33-07:00.
viewCount	number	Views at the time of the run.
likeCount	number	Likes at the time of the run; null when hidden.
category	string	YouTube category, e.g. Music, Education, News & Politics.
isLiveContent	boolean	True for live streams and recorded live streams.
keywords	array	Tags the creator set on the video (may be empty).
description	string	Video description, short (about 300 characters) or full per input; emails and phone numbers are removed.
language	string	Language code of the returned transcript, e.g. en, de-DE, pt-BR.
languageName	string	English name of that language, e.g. German (Germany).
isAutoGenerated	boolean	True when the transcript is YouTube's speech recognition (no punctuation, may misspell names); false when the creator uploaded it.
languageMatched	boolean	False when none of the requested languages existed and the original-language track was returned instead.
translationStatus	string	Values: 'not_requested', 'original' (the video has a native track in translateTo), 'translated' (YouTube's own machine translation), 'unavailable' (YouTube refused it; the original is returned).
translatedFrom		Source language code when translationStatus is 'translated', else null.
availableLanguages	array	Every caption track of the video: code, name, isAutoGenerated. Use it to request another language.
transcript	string	The whole transcript as one text: caption lines joined with spaces. Emails and phone numbers are removed.
segmentCount	number	Number of caption lines.
wordCount	number	Words in transcript, for chunking and token estimates.
segments	array	Caption lines in order: start (seconds from the start), duration (seconds), text. Present when includeSegments is on.
srt		The captions as an SRT subtitle file, when subtitleFormat is srt.
vtt		The captions as a WebVTT subtitle file, when subtitleFormat is vtt.
source		The channel or playlist URL this video came from; null for videos given directly.
scrapedAt	string	When the row was read, ISO 8601.

Error rows, free 3 fields

Error rows, free
Field	Type	What it holds
input		Error rows: the URL or id as given.
code		Error rows: not_found (private, removed, unknown channel), no_match (no captions, or none of the requested type), unavailable (age-restricted, members-only, upcoming, temporary error), blocked (rate limited, retry later), invalid_input.
message		Error rows: what happened, in plain English.

What it can't do

It returns captions YouTube already has. It does not transcribe audio.
It does not return comments or the names of commenters.
Private and members-only videos are not read.

Channel names and URLs are the publisher labels YouTube shows publicly on every video.

What you send

Every field is optional. The example request above fills the ones shown here.

Main input fields
Field	Type	What it does
urls	array	Example: https://www.youtube.com/watch?v=dQw4w9WgXcQ, https://youtu.be/dQw4w9WgXcQ, a Shorts URL, a bare video ID (dQw4w9WgXcQ), a channel (https://www.youtube.com/@kurzgesagt, add /shorts or /streams for those tabs) or a playlist URL. Channels and playlists give up to maxVideosPerSource videos each, channels newest first.
languages	array	ISO language codes in order of preference, e.g. ["en", "de"]; en also matches en-US and en-GB. The first one the video has is used. Empty or none available: the original spoken language (languageMatched is false when none matched).
captionType	string (one of 3)	Values: any (uploaded captions first, else auto-generated), manual (uploaded by the creator only), auto (YouTube speech recognition only). Videos without that type give a free error row. Default: any.
maxVideosPerSource	integer	Transcripts per channel or playlist URL, 1 to 5,000 (default 50). Videos without captions or outside the date range do not count. Direct video URLs are not limited by this.
subtitleFormat	string (one of 3)	Values: none, srt, vtt. srt or vtt adds the captions as a ready subtitle file in the srt or vtt field. Default: none.

More options (7)

More input fields
Field	Type	What it does
translateTo	string	ISO language code, e.g. es. A native track in that language is used when the video has one (translationStatus original). Otherwise YouTube's own machine translation is asked for; YouTube currently refuses it for anonymous clients, so you get the original with translationStatus unavailable.
publishedAfter	string	Date as YYYY-MM-DD, e.g. 2026-01-01. Videos published earlier are skipped for free; on a channel the listing stops at the first older videos.
publishedBefore	string	Date as YYYY-MM-DD, e.g. 2026-06-30. Videos published later are skipped for free.
includeSegments	boolean	Adds segments: one entry per caption line with start and duration in seconds and its text. The joined transcript text is always included.Default `true`
description	string (one of 3)	Values: short (about 300 characters, default), full, none. Emails and phone numbers are removed.Default `"short"`
maxResults	integer	Stops the run after this many transcripts across all URLs, 1 to 1,000,000. A cost cap; leave empty for no limit.
includeErrorRows	boolean	Adds a free row (rowType error, with a code and message) for every video or URL that gave no transcript: no captions, private, age-restricted, not found. On by default.Default `true`

Questions people ask before they top up

What does a run cost?

Each transcript costs $0.0015, which is $1.50 per 1,000. A run that delivers rows also pays a $0.001 start fee.

A run that returns rows costs at least $0.005, start fee included. Failed runs, empty runs and duplicate rows cost $0. You set the most a run may cost, and it stops there.

Do you ever return personal data?

Not about reviewers or users. Reviewer names, profile links, photos, emails and phone numbers are not collected. When a name appears inside a review or a reply, it is replaced with [name].

Job titles and company sizes stay where a site shows them, because they describe a role rather than a person. On YouTube, channel names and URLs are the publisher labels YouTube shows publicly.

How fresh is the data?

Each run reads the live site when you start it, so rows are as current as the site itself.