YouTube Transcript Scraper

Get the transcript of any YouTube video as plain text and timed segments. Give it video URLs or IDs, pick a language, and get back the captions the video actually has, with the title, channel, duration, chapters and a flag telling you whether the captions were written by the creator or generated by YouTube. No API key, no cookies, no browser.

Built for AI pipelines, research and content tools: array input, one item per video, clean error codes instead of failed runs, and SRT or WebVTT when you need subtitle files.

What you get

For every video, one dataset item with:

Field	Description
videoId, url, title, channelName, channelId, durationSeconds, viewCount	Video metadata
text	The full transcript as one string
segments	Timed lines: start, duration, end in seconds and text
language, languageName, captionKind	Which track was returned; captionKind is manual (creator-uploaded) or auto (YouTube speech recognition)
availableLanguages, translatableTo	Every caption track the video offers, and the languages YouTube can auto-translate it into
chapters	Chapters parsed from the description timestamps, when the creator added them
wordCount, segmentCount	Sizes
srt, vtt	Subtitle files as strings, when you switch them on
error, errorMessage	Present only on items that could not be fetched, see below

Sample output

{
  "videoId": "jNQXAC9IVRw",
  "url": "https://www.youtube.com/watch?v=jNQXAC9IVRw",
  "title": "Me at the zoo",
  "channelName": "jawed",
  "channelId": "UC4QobU6STFB0P71PMvOGN5A",
  "durationSeconds": 19,
  "viewCount": 447816985,
  "language": "en",
  "languageName": "English",
  "captionKind": "manual",
  "text": "All right, so here we are, in front of the elephants the cool thing about these guys is that they have really... really really long trunks and that's cool and that's pretty much all there is to say",
  "wordCount": 39,
  "segmentCount": 6,
  "segments": [
    { "start": 1.2, "duration": 2.16, "end": 3.36, "text": "All right, so here we are, in front of the elephants" },
    { "start": 3.36, "duration": 2.4, "end": 5.76, "text": "the cool thing about these guys" }
  ],
  "chapters": [
    { "title": "Intro", "startSeconds": 0 },
    { "title": "The cool thing", "startSeconds": 5 },
    { "title": "End", "startSeconds": 17 }
  ],
  "availableLanguages": [
    { "code": "en", "name": "English", "kind": "manual" },
    { "code": "de", "name": "German", "kind": "manual" }
  ],
  "translatableTo": ["ar", "bs", "de", "es", "fr", "ja"],
  "source": "android",
  "fetchedAt": "2026-11-05T14:47:26.123Z",
  "input": "https://www.youtube.com/watch?v=jNQXAC9IVRw"
}

Input

Field	Type	Default	What it does
videoUrls	array of strings	required	Video URLs or 11-character IDs. Watch, youtu.be, Shorts, embed, live and mobile URLs all work. Duplicates are fetched once.
languages	array of strings	["en"]	Preferred caption languages in order. en also matches en-US and en-GB.
preferManual	boolean	true	Prefer creator-uploaded captions over auto-generated ones in the same language.
fallbackToAnyLanguage	boolean	true	If none of the preferred languages exists, return the video's default track instead of an error.
translateTo	string	none	Ask YouTube to auto-translate the chosen track into this language. Best effort, see below.
includeSegments	boolean	true	Include the timed segments array. Turn off for smaller items.
includeSrt, includeVtt	boolean	false	Add subtitle-file strings.
maxConcurrency	integer	5	Videos fetched in parallel (1 to 20).
proxyConfiguration	object	Apify datacenter proxy	Proxy for all requests.
residentialFallback	boolean	true	If YouTube blocks the datacenter proxy for a video, retry that video through residential proxies. Only used on a detected block.

Minimal input:

{ "videoUrls": ["https://www.youtube.com/watch?v=jNQXAC9IVRw", "dQw4w9WgXcQ"] }

Languages and translation

• The actor returns what the video has. Most videos have an auto-generated track in the spoken language; many also have creator-uploaded tracks in several languages. availableLanguages lists them all so you can make a second request for another language.
• languages is a preference list. With ["bs", "hr", "en"] you get Bosnian if it exists, otherwise Croatian, otherwise English, otherwise the default track (or a LANGUAGE_NOT_AVAILABLE item if you set fallbackToAnyLanguage to false).
• translateTo requests YouTube's own machine translation of the chosen track. YouTube refuses this for some tracks and rate-limits it aggressively, so treat it as a bonus: when it is refused, you still get the original transcript, plus a translationError field saying why. translatableTo tells you which languages YouTube offers for that video.

Errors

The run does not fail because one video is broken. Each problem video is a normal dataset item with an error code and an errorMessage, so your pipeline can branch on it:

error	Meaning	Charged?
INVALID_INPUT	The string is not a YouTube video URL or ID	no
VIDEO_UNAVAILABLE	Deleted, private, region-locked or never existed	yes
NO_CAPTIONS	The video has no caption track at all (live streams, some music videos, captions disabled by the creator)	yes
LANGUAGE_NOT_AVAILABLE	None of your languages exists and fallback is off; the message lists what the video offers	yes
LOGIN_REQUIRED	YouTube only serves this video to signed-in users	yes
BLOCKED	YouTube refused every retrieval path for this video from every proxy tried	no
FETCH_FAILED	Network or parsing failure on every path	no

Items with metadata (title, available languages) are filled in even when the transcript itself could not be returned, so LANGUAGE_NOT_AVAILABLE still tells you which languages to ask for.

Pricing

Pay per event: you pay a small fixed amount per video result (successful transcript, or one of the charged error items above) plus Apify platform usage for the run. Our-side failures (BLOCKED, FETCH_FAILED) and invalid inputs are never charged. Set a maximum total charge on the run and the actor stops cleanly when it is reached, with the unprocessed count in the status message.

Typical cost: a run of 100 videos takes about a minute on 512 MB of memory and transfers roughly 0.1 MB per video.

Reliability

YouTube changed caption delivery in 2025 and 2026: the URLs most scrapers use now return HTTP 200 with an empty body unless a proof-of-origin token is attached, and datacenter IP ranges are throttled. This actor:

• fetches captions through three independent paths (two mobile InnerTube clients, then the web player with a minted proof-of-origin token) and moves to the next one on any block or empty response;
• checks every response at body level, so an empty 200 is treated as a failure, never saved as an empty transcript;
• retries a blocked video once through residential proxies when residentialFallback is on;
• is tested daily with the default input.

source on each item tells you which path produced it, which is useful when reporting an issue.

Limitations

• Live streams in progress have no caption track; the recording usually gets one a few hours after the stream ends.
• Videos with captions disabled by the creator, and most music videos, return NO_CAPTIONS.
• Private, members-only and age-gated videos that require a signed-in account are not supported; the actor never uses cookies.
• Transcript timing comes from YouTube's caption data. Auto-generated captions contain YouTube's own markers such as [music] or >> speaker changes.
• Chapters are parsed from description timestamps; videos whose chapters were set only in YouTube Studio without timestamps in the description return an empty array.

Using it from code

Apify API, Node:

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor('serentity/youtube-transcript-scraper').call({
    videoUrls: ['https://www.youtube.com/watch?v=jNQXAC9IVRw'],
    languages: ['en'],
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items[0].text);

Python:

from apify_client import ApifyClient
client = ApifyClient("YOUR_TOKEN")
run = client.actor("serentity/youtube-transcript-scraper").call(run_input={"videoUrls": ["jNQXAC9IVRw"]})
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item["title"], item.get("wordCount"), item.get("error"))

The actor also works as a tool through the Apify MCP server: give an agent the list of URLs and it gets back the transcript text and segments in one call.

Legal

Captions are public data that YouTube serves to any viewer. You are responsible for using transcripts in line with YouTube's terms of service and the copyright of the video's creator. The actor does not download video or audio, does not sign in, and does not collect personal data beyond the public channel name and ID shown on the video page.