YouTube Transcript Scraper & API · $0.50/1K avatar

YouTube Transcript Scraper & API · $0.50/1K

Pricing

from $0.30 / 1,000 transcripts

Go to Apify Store
YouTube Transcript Scraper & API · $0.50/1K

YouTube Transcript Scraper & API · $0.50/1K

Get YouTube transcripts and captions in bulk from video, Shorts and live links: full text, timestamped lines, the language you pick from all available captions, plus title, channel, views and date. No API key, no login. $0.50 per 1,000 transcripts. JSON, CSV, Excel.

Pricing

from $0.30 / 1,000 transcripts

Rating

0.0

(0)

Developer

Samy

Samy

Maintained by Community

Actor stats

0

Bookmarked

3

Total users

2

Monthly active users

6.3 days

Issues response

17 days ago

Last modified

Features

Limited permissions

Share

YouTube Transcript Scraper & API

YouTube Transcript Scraper extracts public captions from video, Shorts, channel and playlist links, and keyword searches. Get full text, timestamped lines, video metadata, subtitles and citation chunks for $0.50 per 1,000 delivered videos on the Free tier, plus the run-start fee. Apify's $5 monthly free credits cover about 9,990 transcripts in ten batch runs, assuming no other credit use.

No YouTube login or YouTube API key is required. Export JSON, CSV or Excel, read TXT, SRT, VTT or Markdown, or call the Actor through the API, n8n, Make or MCP.

  • Full caption text with title, channel, description, views and exact publish date when YouTube provides it.
  • Ordered language preferences, strict language matching, manual or automatic caption selection, and track discovery.
  • Native caption archives and optional YouTube machine translation (best effort), without extra transcript events.
  • Readable paragraphs, description chapters, bounded AI chunks and timestamped keyword hits.
  • A per-video delivery report, with one dataset row and one transcript charge for each delivered video.

What can YouTube Transcript Scraper do?

Paste video URLs or IDs into Video links or IDs. For channels or playlists, use Channels and playlists. A channel can supply its Videos, Shorts or Live tab. Archived streams work when their captions are available. An active live stream or an upcoming premiere has no complete transcript yet.

Use searchQueries to discover videos through YouTube search. Each query has its own maxVideosPerSource cap. Search supports uploadDate of hour, today, week, month or year; searchType of videos or shorts; duration of under3, under4, 3to20, 4to20 or over20; and searchSort of relevance or popularity. features accepts live, 4k, hd, subtitles, creativeCommons, 360, vr180, 3d, hdr and location. searchLanguage and searchCountry select the search locale. YouTube ignores upload-date and rating sorts, so those sorts are unavailable. Search filters affect discovery; they do not guarantee readable captions.

The default source cap is 50 considered videos per channel, playlist or search query. A channel shares that cap across the selected tabs, in their listed order. Explicit tab URLs choose that tab. Playlists keep YouTube's order and are not filtered by the channel-tab setting.

A video found through several sources is exported once per run. With includeProvenance: true, the row's source identifies the source that exported it. RUN_SUMMARY.delivery.results[].sources retains every observed association. Resurrecting the same run restores its saved progress and exported-video identities.

language stays the first choice, with languages tried next. Prefer mode falls back to available captions and reports the actual language. Strict mode reports no_captions when none of the requested languages matches. Regional variants match their primary language, so en accepts en-GB.

allLanguages keeps the primary transcript at the top level and returns extra native tracks in tracks[]. maxTracks, default 5, counts the primary plus those extra tracks. listTracksOnly downloads no caption text and puts available-track metadata in the delivery report. It creates no charged transcript rows.

translateTo requests YouTube's own machine translation. Source text stays in transcript; successful translations have their own text, segments and requested files. YouTube often refuses these translation requests while the original captions still work: in our October 2026 tests it refused most of them, so treat translation as best effort. A refused target has translation_unavailable, empty translated text and no extra item fee. A target equal to the source language is labelled source_language.

Get only new videos on a schedule

Set onlyNewVideos: true to skip videos delivered by earlier runs of the same watchlist. The first run exports matching videos, including older videos within your filters. Each exported video creates a delivery record in a persistent named key-value store in your Apify account. A skipped video has no new item or add-on charge. The run-start fee still applies.

Omit watchlistName to use a name derived from the material input, including normalized sources and output choices. Language, translation, subtitle format and output-filter changes select a new default history. A relative date such as 7 days keeps its identity as the schedule advances. An explicit watchlistName keeps one history when you change the input. Use separate names for independent watches and for the two Actors. Names contain 3 to 63 letters, digits or hyphens and start and end with a letter or digit. Choose a new name to start a fresh history.

Run schedules without overlap. The named store does not provide an atomic claim, so two overlapping runs can both export and charge for the same video. A crash after export and before the delivery-record write can also repeat a video in a fresh run. Resuming the original run replays its dataset receipts to repair that gap. This feature does not promise exactly-once delivery across runs. A store read failure stops new work; a record write failure stops further exports after counting the rows already delivered.

Videos without an exported primary row remain eligible next time, including unavailable or captionless videos. Optional add-ons do not refresh an already delivered video. History persists until you choose another name or remove the named store yourself. Use the daily-new-videos saved input as a starting point, then add it to your Apify schedule.

Transcript source caps count considered videos, including already delivered ones. Increase maxVideosPerSource when a channel or search can gain more videos between scheduled runs than the cap. Discovery-only, date-filtered and keyword-filtered videos do not create delivery records.

What data can you get from YouTube?

FieldsContents
videoId, url, title, thumbnailUrlVideo identity and basic metadata.
channelName, channelId, channelHandle, channelUrlPublishing channel. A missing handle is null.
durationSeconds, viewCount, publishedAt, category, descriptionVideo details. Hidden counts or unavailable metadata are null. Dates use UTC.
language, languageName, isAutoGeneratedThe primary caption track actually returned.
availableLanguagesThe existing {code, name, isAutoGenerated} list, kept in its original shape.
availableTracks, tracksPublic track IDs and default flags, plus any additional native transcripts requested. No signed caption-download URLs.
transcript, segments, scrapedAtFull primary text, optional raw timed lines, and retrieval time.
files, translationsRequested representations and YouTube translation results.
paragraphs, chunks, chapters, matchesReadable passages, citation windows, description timestamps and keyword occurrences.
source, provenance, artifactsSource association, caption coverage information and links to large optional output.

Legacy inputs keep the original 20 public fields, including segments by default. New fields are omitted until requested. Set includeParagraphs for paragraphs, includeChapters for chapters, includeProvenance for source and coverage details, or includeAvailableTracks for track identifiers. These flags default to false. Set chunkSize or chunkOverlap to enable chunks. allLanguages, translateTo, outputFormats, and keywords enable only their respective tracks, translations, files, and matches. artifacts appears only when requested output must move to storage.

Paragraphs merge caption lines by pauses and punctuation. Chapters come from timestamps in the description, such as 0:00 Introduction. Paragraphs and chunks identify the chapter active at their start.

Chunks use approximately four Unicode code points per token. chunkSize defaults to a target of 500; chunkOverlap defaults to zero. Each chunk has a stable ID, start and end, an approximate token count and a video seek URL. Chunks end on complete-word boundaries and never split a grapheme. A word larger than the target stays intact and can exceed the target. Start times use the first included caption line and never decrease. Seek URLs use the whole second at or before that start. Overlap adjusts to complete-word boundaries. Only verified repeated JSON3 display lines are removed from derived output; legacy text and raw segments retain every spoken word.

The dataset's Overview view shows one video per row. AI chunks expands the nested chunks for reading and export. That view does not create extra charged dataset items.

Optional representations that would take a row over 8 MB move into a JSON key-value-store record linked by artifacts.url. That record contains the full files, paragraphs, chunks, chapters, matches, tracks and translations. The legacy primary transcript and raw segments remain in the dataset. If the primary row alone is too large, the video fails with output_too_large and has no transcript charge. Turning off includeSegments can reduce that primary row.

How to scrape YouTube transcripts

  1. Open the Actor in Apify Console.
  2. Paste your video links or IDs, or enter channel, playlist or search sources. Remove the sample video links for a source-only run.
  3. Choose a preferred language. Use Strict mode if a fallback language would be unsuitable.
  4. Select any subtitle formats, translation targets, keywords or date bounds you need.
  5. Set a spending limit and click Start.
  6. Open Delivered transcripts and Delivery report and available tracks in the run output. Use JSON when you need nested tracks, translations or chunks.

The old videos, language, includeSegments and proxy inputs keep their behavior. Existing output fields retain their names and types. includeSegments: false still omits the raw segments array.

How much does it cost to scrape YouTube transcripts?

Apify discount tierCommon planPer 1,000 delivered videos
FREEFree$0.50
BRONZEStarter$0.40
SILVERScale$0.35
GOLDBusiness$0.30
PLATINUMEnterprise$0.30
DIAMONDEnterprise$0.30

Apify assigns your actual discount tier. Check the Actor's Prices tab before running.

The run-start event is $0.0002 per GB of memory, with at least one event. Default runs use 512 MB, so a run pays one $0.0002 start. At the Free tier, one delivered video in its own run costs $0.0007. One run delivering 1,000 videos costs $0.5002.

Only exported transcript rows have the transcript item fee. Captionless, unavailable, missing, failed, filtered and discovery-only videos have no transcript item fee. Extra native tracks, YouTube translation, paragraphs, chunks, chapters, keyword hits and formats do not add transcript events. The start fee still applies to a run that exports nothing.

Is scraping YouTube transcripts free? Apify's free plan includes $5 in monthly credits. At $0.0005 per delivered video, $5 covers about 9,990 transcripts in ten batch runs, or about 7,140 when each run delivers one video. This assumes no other credit use. Using your own proxy service can incur that service's charges.

The three high-adoption transcript Actors inspected in the 2026-10-30 research listed native-caption events of $2.75, $5 and $10 per 1,000 before starts and add-ons. Other Store Actors can be cheaper. No competitor coverage or latency benchmark is claimed here.

Input example

{
"videos": ["https://www.youtube.com/watch?v=iG9CE55wbtY"],
"language": "en",
"includeSegments": true,
"includeParagraphs": true,
"includeChapters": true,
"includeProvenance": true,
"includeAvailableTracks": true,
"outputFormats": ["srt", "vtt"],
"translateTo": ["es"],
"chunkSize": 500,
"chunkOverlap": 50
}

For a bounded channel scan, replace videos with "channels": ["@TED"] and add "maxVideosPerSource": 10.

InputDefault and purpose
searchQueriesOptional keyword queries. Each query has its own source cap. Blank queries are skipped and repeated queries are deduplicated without regard to case.
uploadDate, searchType, duration, features, searchSortOptional YouTube discovery filters described above. Default relevance order.
searchLanguage, searchCountryOptional search locale. These fields do not select caption languages.
onlyNewVideosDefault false. Skip earlier deliveries from this watchlist. Schedule without overlapping runs.
watchlistNameOptional named store. Otherwise the Actor hashes material input to select its history.
videosVideo links or IDs, up to 10,000 entries. Provide videos, sources, or both.
channelsChannel and playlist sources, up to 100 entries.
contentTypes["videos"]. Channel tabs, in scan order. Explicit tab URLs override this list.
maxVideosPerSource50. Scan cap from 1 to 5,000, including failed, captionless, duplicate and filtered videos.
publishedAfter, publishedBeforeNo bounds. Exclusive ISO date or timestamp limits, within the scan cap. Date-only values mean midnight UTC.
language, languages"en", then no extra priorities. Native caption selection, with regional matching.
captionType, languageMode"any", "prefer". Use manual or auto, and strict to prevent language fallback.
allLanguages, maxTracksfalse, 5. Primary plus at most four extra native tracks by default; maximum ten tracks.
listTracksOnlyfalse. Track metadata in the delivery report, with no caption downloads or dataset rows.
translateToNo targets. Up to three language codes; YouTube machine translation is best effort.
outputFormatsNo formats. Any of txt, srt, vtt, md and json.
includeSegmentstrue. Include the raw timed caption lines.
chunkSize, chunkOverlapOmitted by default. Set either to enable chunks, with targets of 500, 0 when unspecified. Overlap must be smaller than chunk size.
keywords, onlyVideosWithMatchesNo keywords, false. Literal, case-insensitive phrase matching and optional export filtering.
includeParagraphs, includeChapters, includeProvenance, includeAvailableTracksfalse. Opt in to the corresponding optional row fields.
proxyAutomatic routing. Enter your own proxy URLs only when you want your proxy service used.

A publish-date bound excludes a video whose exact date cannot be read. Keywords apply to the primary transcript, including phrases across adjacent caption lines. A filter-only run reports excluded videos as filtered, with outside_publish_range, publish_date_unknown or no_keyword_matches.

Output example

This is trimmed real output from a local run on 2026-10-30. Description, transcript, files and arrays are excerpted. That translation attempt was refused; the original transcript and requested English subtitles still delivered. This extended example requests paragraphs, chapters, chunks, provenance, available tracks, translations, and file formats. Default rows contain only the legacy fields.

{
"videoId": "iG9CE55wbtY",
"url": "https://www.youtube.com/watch?v=iG9CE55wbtY",
"title": "Do schools kill creativity? | Sir Ken Robinson | TED",
"channelName": "TED",
"channelId": "UCAuUUnT6oDeKwE6v1NGQxug",
"channelHandle": "@TED",
"channelUrl": "https://www.youtube.com/@TED",
"durationSeconds": 1203,
"viewCount": 25077478,
"publishedAt": "2007-01-07T05:00:34.000Z",
"thumbnailUrl": "https://i.ytimg.com/vi/iG9CE55wbtY/hqdefault.jpg",
"category": "Education",
"language": "en",
"languageName": "English",
"isAutoGenerated": false,
"source": "https://www.youtube.com/watch?v=iG9CE55wbtY",
"scrapedAt": "2026-10-30T18:26:12.789Z",
"description": "Visit http://TED.com to get our entire library of TED Talks, transcripts, translations, personalized talk recommendations and more.\n\nSir Ken Robinson makes an e",
"availableLanguages": [
{
"code": "en",
"name": "English",
"isAutoGenerated": false
},
{
"code": "en",
"name": "English (auto-generated)",
"isAutoGenerated": true
},
{
"code": "es",
"name": "Spanish",
"isAutoGenerated": false
}
],
"transcript": "Good morning. How are you? (Audience) Good. It's been great, hasn't it? I've been blown away by the whole thing. In fact, I'm leaving. (Laughter) There have been three themes running through the conference, which are relevant to what I want to talk about. One is the extraordinary evidence of human creativity in all of the presentations that we've had and in all of the people here; just the variety of it and the range of it. The second is that it'",
"segments": [
{
"start": 27.103,
"duration": 2.575,
"text": "Good morning. How are you?"
},
{
"start": 29.702,
"duration": 1.403,
"text": "(Audience) Good."
},
{
"start": 31.129,
"duration": 1.668,
"text": "It's been great, hasn't it?"
}
],
"files": {
"srt": "1\n00:00:27,103 --> 00:00:29,678\nGood morning. How are you?\n\n2\n00:00:29,702 --> 00:00:31,105\n(Audience) Good.\n\n3\n00:00:31,129 --> 00:00:32,797\nIt's been great, h"
},
"translations": [
{
"language": "es",
"status": "translation_unavailable",
"text": "",
"segments": [],
"files": {}
}
],
"chapters": [],
"provenance": {
"transcriptSource": "youtube_captions",
"languageFallback": false,
"coveredUntilSeconds": 1165.213,
"completeness": "not_verified"
}
}

RUN_SUMMARY retains the existing run report and adds delivery. Its requested count equals the number of unique considered videos. Every video has one outcome, a stable code, its source associations and a charged flag.

OutcomeMeaning
deliveredA transcript row is present in the dataset.
no_captionsNo accessible captions match the request, including language or caption-type restrictions.
unavailableYouTube reports a private, removed or otherwise unavailable video.
not_foundYouTube reports a missing video or returns a bare not-found response.
failedThe transcript could not be delivered. Inspect the code and try again later.
filteredAn explicit date or keyword filter excluded the video. No item charge.
tracks_listedDiscovery-only metadata was returned in the summary. No item charge.

The last two outcomes apply to the corresponding opt-in modes. A normal extraction has the first five outcomes. The final run status gives their counts and failure reason codes.

Local checks on 2026-10-30 delivered the sample TED talk, a fresh TED channel upload and the 8,752-second micrograd lecture from its playlist. The lecture output included 22 description chapters and 62 default-sized chunks. Its last caption started at 8,750.8 seconds. These checks establish those examples only. Caption coverage is reported in provenance; YouTube captions alone cannot independently verify every spoken word.

Who uses YouTube transcripts?

AI and RAG builders

Index chunks with their stable IDs and video seek links. Keep language and isAutoGenerated with each passage. Use the reported token estimate to size a first pass, then apply your model's tokenizer if exact limits matter.

Researchers and journalists

Export captions with metadata and source dates. Use literal phrase hits to locate passages, and follow the timestamped URLs to check quotations against the video.

Content and localization teams

Download native language tracks and SRT or VTT contents. Use translateTo when YouTube machine translation is suitable. Review automatic captions and translations before publication.

Use the API, n8n, Make or MCP

You need an Apify token for the Apify API. You do not need a YouTube API key. Keep the Apify token in your environment.

JavaScript with built-in fetch:

const response = await fetch(
"https://api.apify.com/v2/acts/sones~youtube-transcript-scraper-lowcost/run-sync-get-dataset-items?clean=1",
{
method: "POST",
headers: {
Authorization: `Bearer ${process.env.APIFY_TOKEN}`,
"Content-Type": "application/json",
},
body: JSON.stringify({ videos: ["iG9CE55wbtY"], language: "en" }),
},
);
if (!response.ok) throw new Error(`Apify returned HTTP ${response.status}`);
const transcripts = await response.json();

Python with the standard library:

import json
import os
import urllib.request
request = urllib.request.Request(
"https://api.apify.com/v2/acts/sones~youtube-transcript-scraper-lowcost/run-sync-get-dataset-items?clean=1",
data=json.dumps({"videos": ["iG9CE55wbtY"], "language": "en"}).encode(),
headers={
"Authorization": f"Bearer {os.environ['APIFY_TOKEN']}",
"Content-Type": "application/json",
},
method="POST",
)
with urllib.request.urlopen(request, timeout=300) as response:
transcripts = json.load(response)

For long channel runs, track discovery or detailed failure accounting, start an asynchronous run through /v2/acts/sones~youtube-transcript-scraper-lowcost/runs. After it finishes, read its default dataset and the RUN_SUMMARY record from its default key-value store. An empty dataset can mean no captions, a strict-language miss, an explicit filter or discovery-only mode. The summary identifies which occurred.

In n8n, use the Apify integration or an HTTP Request node to submit the same JSON. Wait for the run to finish, read its dataset, then map transcript, chunks or matches to your next node. Read the delivery report for skipped or failed videos.

In Make, run the Actor with the Apify integration, wait for completion and retrieve dataset items. Use an iterator for nested chunks or translations. Read the key-value-store summary for discovery-only output.

For MCP, configure the Apify MCP server in your agent client. Enable this Actor and supply the same input fields. Ask the agent to inspect the delivery report before treating an empty dataset as an empty transcript.

Ready-made tasks

The release includes ten saved-input templates in tasks.json. Public example-task pages have not been published. Copy these starting inputs into the Actor's JSON editor, or save them as tasks in your account. Set a spending limit when saving a task.

TemplateStarting input
One video with timestamps{"videos":["iG9CE55wbtY"],"language":"en","includeSegments":true}
Bulk transcripts{"videos":["iG9CE55wbtY","dQw4w9WgXcQ"],"language":"en"}
Channel transcripts for LLMs{"channels":["@TED"],"maxVideosPerSource":25,"outputFormats":["md","json"],"chunkSize":500,"chunkOverlap":50}
Spanish playlist subtitles{"channels":["PLAqhIrjkxbuWI23v9cThsA9GvCAUhRvKZ"],"maxVideosPerSource":5,"language":"es","languageMode":"strict","outputFormats":["srt","vtt"]}
A Shorts transcript{"videos":["https://www.youtube.com/shorts/R6yNUnRXZ64"],"outputFormats":["txt"]}
Caption language audit{"videos":["iG9CE55wbtY"],"listTracksOnly":true}
Multilingual archive{"videos":["iG9CE55wbtY"],"allLanguages":true,"maxTracks":5,"outputFormats":["txt","srt","json"]}
Channel keyword research{"channels":["@3blue1brown"],"maxVideosPerSource":10,"keywords":["neural network","gradient"],"onlyVideosWithMatches":true}
YouTube caption translation{"videos":["iG9CE55wbtY"],"translateTo":["es","fr"],"outputFormats":["srt","vtt"]}
Long lecture for RAG{"videos":["VMj-3S1tku0"],"outputFormats":["md","txt"],"chunkSize":500,"chunkOverlap":50}

FAQ

Is there a YouTube transcript API without a YouTube API key?

Yes. This Actor reads public caption tracks without a YouTube API key or login. API calls to Apify require an Apify token. Private, members-only and age-restricted content is not made accessible by this Actor.

Is this the cheapest YouTube transcript API on Apify?

No lowest-price claim is made. Some smaller Actors list lower transcript fees. Compare your plan's delivered-video rate, run-start fee, native caption coverage and the output you need. Small one-video runs cost more per video because of the start fee.

Can I get other languages or auto-generated captions?

Yes, when YouTube exposes them. Set captionType to manual, auto or any. Use Strict mode to prevent a language fallback. availableLanguages lists native tracks; translateTo requests machine translation and can be refused independently of the original captions.

What happens to videos without captions?

They get a no_captions outcome and a reason code in the delivery report. No transcript row is exported or charged. The run-start fee applies. This Actor does not download audio or run speech recognition to invent missing captions.

Can I get only new videos on a schedule?

Yes. Enable onlyNewVideos, optionally set watchlistName, and schedule runs without overlap. See Get only new videos on a schedule for persistence, source caps and the export/write crash limit.

Why is my transcript in another language?

Prefer mode found no track in your requested language list. The row reports its actual language. With includeProvenance: true, it also reports provenance.languageFallback. Choose Strict mode to receive a no_captions outcome instead.

Does the Actor return every spoken word?

It returns the caption text YouTube supplies. Closed but incomplete upstream captions, uncaptioned speech, music and silence can leave gaps. Truncated XML responses are rejected. With includeProvenance: true, the last caption end is exposed in provenance.coveredUntilSeconds; completeness remains not_verified.

Captions and video metadata can be protected by copyright and YouTube's terms. Use public material you are entitled to process, respect creators' rights and check the rules that apply to your use. The Actor does not provide a license to republish a transcript.

How do I get help or request a feature?

Open the Actor's Issues tab with the run ID, affected public video URLs, requested language and delivery reason codes. Remove proxy credentials and API tokens. No response-time guarantee is advertised.

More YouTube scrapers by sones

Use YouTube Channel & Video Scraper when you need channel or playlist video listings and exact metadata without transcripts. This transcript Actor expands channels and playlists itself, so native caption extraction does not require chaining the two Actors.

YouTube's uploadDate filter narrows search discovery. The Actor also enforces the window locally from tile or player dates. hour, today, week, month, and year mean the last hour, 24 hours, 7 days, 30 days, and 365 days. A video with no usable date is excluded.