YouTube Transcript Scraper — full transcripts for a whole channel

Paste one YouTube channel or playlist URL and get up to 500 transcripts in a single run: the full text of every video, timestamped segments, the subtitle language, the publish label, the duration and the channel fields — one flat dataset row per video.

Built for content and SEO teams archiving competitor channels, and for anyone loading a video library into a RAG index or a search corpus. One delivered video is one billed row; videos you already pulled in a previous run of the same source are skipped for free.

Run it from the Apify Console, the API, the CLI, or through the Apify MCP server (mcp.apify.com) so an AI assistant can fetch a channel's transcripts on its own.

What does the YouTube Transcript Scraper do?

It expands one source — a channel handle, a channel ID, a channel URL or a playlist URL — into its long-form videos, then fetches the subtitles of each video in the language you asked for. Output is a flat table: one row per video, 18 fixed columns, carrying both the plain-text transcript and the timestamped segments behind it. You can also skip expansion entirely and hand it an explicit list of video IDs or watch URLs.

It does not summarise, translate or rewrite anything. It answers one question per run: what was said in these videos, and when.

Why use the YouTube Transcript Scraper

• One URL instead of 500 runs. Up to 500 videos per run from a single channel or playlist link, 100 by default. No separate step to build the video-ID list.
• Text and timestamps in the same row. transcriptText for search and embeddings, segments[] with start, duration and end in seconds for citations and clipping.
• Incremental re-runs by date. Set Published after to your last run's date and the run delivers only videos you do not already have — repeats are skipped without a fetch and without a charge.
• 28 subtitle languages, honestly reported. transcriptLanguage echoes the language actually returned, not the one you requested, so a silent fallback to another language shows up in the data instead of hiding inside a "successful" row.
• Missing subtitles are an answer, not a crash. A video with no subtitles in your language comes back as a row with notFound: true and an empty transcript, so a 500-video job never dies on video 137.
• Free cache, free repeats. A video fetched in the last 24 hours is re-delivered from cache at no charge, with its original finishedAt.

How to scrape YouTube transcripts from a channel

1. Put the channel or playlist in Channel or playlist — @McCannDogs, a UC… channel ID, a channel URL or a playlist URL.
2. Set Maximum videos per run (default 100, maximum 500). One returned video is one row.
3. Pick the Subtitle language — English by default, 28 languages in the dropdown.
4. For a weekly top-up, set Published after to the date of your previous run. Leave it empty to pull the full archive.
5. Optionally set a Country if you want the data requested from a specific market.
6. Run the Actor — rows stream into the dataset as each video finishes. Open the Overview view for a readable table, then export to JSON, CSV or Excel, or read the rows from the API.

Minimal input:

{ "sourceUrl": "@McCannDogs", "maxVideos": 100 }

Incremental Spanish top-up:

{ "sourceUrl": "@McCannDogs", "publishedAfter": "2026-07-01", "language": "es" }

{ "videoIds": ["dQw4w9WgXcQ", "https://www.youtube.com/watch?v=UF8uR6Z6KLc"] }

What transcript data you get per video

One dataset row = one video, always the same 18 keys. Example row (segment list truncated for space):

{
  "videoId": "dQw4w9WgXcQ",
  "url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
  "title": "How to teach loose lead walking",
  "channelId": "UC8dGqrQ6cQoT1nZ2sDdZAOA",
  "channelHandle": "@McCannDogs",
  "channelTitle": "McCann Dog Training",
  "publishedAt": null,
  "publishedText": "3 days ago",
  "durationSec": 1104,
  "transcriptLanguage": "en",
  "transcriptText": "Hey there, welcome back to the channel. Today we are working on loose lead walking ...",
  "segments": [
    { "text": "Hey there, welcome back to the channel.", "start": 0.16, "duration": 3.2, "end": 3.36 },
    { "text": "Today we are working on ...", "start": 3.36, "duration": 2.72, "end": 6.08 }
  ],
  "segmentCount": 61,
  "wordCount": 1873,
  "notFound": false,
  "source": "primary",
  "contentHash": "3f0a1c7d5b9e2a64c8d1f0b7e6a4539c2d81ff40a7b6c5d4e3f2a1b0c9d8e7f6",
  "finishedAt": "2026-07-29T09:14:02.371Z"
}

The Issues dataset view narrows a run down to the rows worth a second look: notFound: true and source: "fallback".

How incremental re-runs work

Re-running the same channel with Published after set to your previous run's date does two things. First, the date filter drops older videos before anything is fetched. Second, videos already delivered and billed for this source in this language are skipped even when the relative publish label ("1 month ago") still passes the filter — the run keeps a record of what it already handed you, so a weekly top-up returns new videos only, and the log states how many were skipped and why.

Without Published after nothing is de-duplicated: a run with no date threshold is a request for the full archive, and quietly dropping rows would leave holes in a corpus you paid for.

Input parameters

How much does the YouTube Transcript Scraper cost?

Pay-per-event billing: you pay per delivered row, never per request and never per channel.

Minimum run budget: $0.014 — one run start plus one row at the Free-plan price. A 500-video run on the Free plan is $0.005 + 500 × $0.009.

You are never charged for cache hits (a video fetched within 24 hours is re-delivered free), for repeats skipped in incremental mode, for failures (no row, no charge), or for retries and backup paths inside a row — those are included in the per-row price. A notFound video is a delivered answer and bills as one row. A run with nothing to process does not even bill the run start, and when the remaining run budget drops below the minimum the run stops before its first request instead of half-delivering.

YouTube Transcript Scraper vs single-video transcript tools

What this actor does not do

• It does not summarise, translate or clean transcripts — you get the text as published.
• It does not generate subtitles for videos that have none; those rows come back notFound.
• It does not invent a publishedAt when the source lists only "4 years ago" — the raw label is reported instead of a made-up date.
• It does not include Shorts, comments, chapters, likes or view counts in this version.
• It does not download audio or video files, or touch members-only and private videos.
• It does not go past 500 videos in one run — split larger archives by date across runs.

FAQ

How many transcripts can I get in one run?

Up to 500 videos per run, 100 by default. For a bigger archive, run the same channel several times with different Published after dates.

Why is publishedAt null while publishedText says "3 days ago"?

Because the channel listing carries a relative label, not a date. Writing "3 days ago" into an ISO field would invent a precision the source never gave. The filter uses the label internally, and the raw label stays in publishedText so you can see exactly what was published.

What happens to videos without subtitles?

They come back as normal rows with notFound: true, an empty transcriptText and segments: []. They are delivered answers and bill as one row each, so a channel that captions nothing shows up in your data instead of failing the run.

I asked for Spanish but got English — why?

Some videos only carry subtitles in another language. transcriptLanguage always reports the language actually returned, so you can filter or re-queue those rows. It is reported, never masked.

Do cache hits and skipped repeats cost anything?

No. A video fetched in the last 24 hours is re-delivered from cache, and in incremental mode a video an earlier run already delivered is not fetched at all. Neither is charged.

Can I run it from an AI agent or on a schedule?

Yes. Use the Apify API, the CLI, or the Apify MCP server (mcp.apify.com), and schedule weekly runs with Published after for a transcript corpus that maintains itself.

Related actors

• YouTube Research Digest — when you want an AI digest of what the top videos for a keyword say, rather than the raw transcripts of one channel.
• YouTube Influencer Finder — find the channels worth transcribing, with subscriber stats and social profiles.

Legal & support

• This actor collects publicly available video information: the video list a channel or playlist publishes, and the subtitle tracks published with those videos. It does not access private, members-only or log-in-gated content, and downloads no audio or video files.
• Transcripts are returned verbatim, with the timestamps as published — nothing is rewritten, translated or inferred, so you can verify every line against the source video.
• Transcripts and video metadata are third-party content. You remain responsible for how you use them, including copyright, YouTube's terms and any quoting, attribution or storage rules that apply to you — especially when transcripts feed a public product or an AI index.
• Output is provided as-is for research, archiving and internal analysis; subtitle tracks are point-in-time content that the publisher can change or remove at any time.
• Questions or a bug? Open an issue on the Actor's Issues tab, or see the API tab.

SEO keywords

youtube transcript scraper, youtube transcript api, youtube transcript tool, download youtube transcripts, youtube channel transcripts, bulk youtube transcripts, youtube subtitles scraper, youtube captions api, extract youtube transcript, youtube playlist transcripts, video transcript dataset, youtube transcript with timestamps, scrape youtube subtitles, transcripts for rag, youtube video text extraction, youtube transcript export csv, youtube transcript json, multilingual youtube transcripts, incremental youtube scraping, youtube content analysis.

Built by doesaiknow.