YouTube Transcript Scraper - Captions & Subtitles

YouTube Transcript Scraper extracts the full transcript of any public YouTube video, playlist or channel as clean text with timestamps. Get YouTube subtitles and captions, human-written or auto-generated, in any language the video offers, plus title, channel, views and publish date. YouTube to text for AI, RAG and research. No API key.

What you get

One dataset row per video transcript, ready to export as JSON, CSV or Excel.

{
  "video_id": "dQw4w9WgXcQ",
  "url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
  "title": "Rick Astley - Never Gonna Give You Up (Official Video) (4K Remaster)",
  "channel_name": "Rick Astley",
  "duration_seconds": 213,
  "published_at": "2009-10-24T23:57:33-07:00",
  "view_count": 1793953812,
  "language": "en",
  "is_generated": false,
  "transcript_text": "[♪♪♪]\n♪ We're no strangers to love ♪\n♪ You know the rules and so do I ♪",
  "scraped_at": "2026-07-18T09:03:17.597050+00:00",
  "_meta": {"author": "https://apify.com/herus13", "status": "ok", "message": null}
}

• transcript_text is the whole transcript as one string, one caption line per line: the shape an LLM prompt, a summarizer or a search index wants.
• transcript_segments is the same transcript with timestamps, as a JSON list of {text, start, duration} (seconds). Use it for jump-to-moment links, quotes with a time, or .srt and .vtt subtitle files.
• language is the caption track you got, and is_generated says whether it was auto-generated by speech recognition (true) or written by a person (false).
• available_languages lists every caption track the video offers, so you can see what else to ask for.
• Rows also carry channel_id, description and thumbnail_url.

_meta is run metadata, not scraped data. A video that yields no transcript appears as a row with only _meta (status failed, the reason in message), and those rows are never charged. Every other row is a delivered transcript.

What it costs

Event	Price
Transcript (transcript)	$0.003

You pay per transcript delivered, not per minute: $0.003 a video, with no fee to start the run. A hundred videos is about $0.30. A video with no captions, or none in the language you asked for, returns only a _meta row, is named in the run log and is not charged. Expanding a playlist or channel into its videos costs nothing, so you only pay for what comes back. The residential proxy the run needs is included in that price.

Input

Field	Type	Required	Default	What it does
video_urls	array	no	—	YouTube links to transcribe, one per entry, mixing three kinds freely. A video: watch?v=…, youtu.be/…, /shorts/…, /embed/…, or a bare 11-character ID. A playlist: /playlist?list=PL… — every video in it is transcribed. A channel: /@handle, /channel/UC…, /c/Name or /user/Name — its uploads, newest first. A watch?v=…&list=… link counts as the single video. Duplicates are dropped, and an entry that is not a YouTube link or ID is skipped. Leave empty and nothing runs.
language	string	no	—	Caption language as a language code, for example en, es, zh, ja, pt-BR. A short code also matches its regional tracks: zh takes zh-CN or zh-TW. Leave empty to take the track the video treats as its default. Captions come only in languages the video has a track for; nothing is machine-translated. A video with no track in that language returns only a _meta row, is named in the run log, and is not charged. Each row's available_languages lists every track the video offers.
prefer_manual	boolean	no	true	When a video has both human-written captions and auto-generated (speech-recognition) captions in the language you asked for, take the human ones, which are punctuated and more accurate. Default true. Set it to false to take the auto-generated track instead. A video with only one kind of track returns that kind regardless. Each row's is_generated field says which kind you got.
max_videos_per_source	integer	no	—	How many videos to take from each playlist or channel you paste, at least 1. Leave empty to use 100, the built-in cap that stops one channel link from turning into a very large run. It does nothing to entries that are already a single video.
max_records	integer	no	—	Ceiling on how many videos the whole run processes, counted after playlists and channels have expanded, at least 1. Leave empty for no ceiling. It is the cost ceiling for a run that starts from a channel whose size you do not know.
proxyUrls	array	no	—	Leave empty and the run uses the residential proxy this actor ships with, included in the price of the run. To route the run through your own account instead, add one gateway URL per entry, for example http://user:pass@host:port — works with any provider that issues URLs. When set, the run uses these first and switches to the included proxy if they stop working, so a broken proxy never fails the run.

How to run it

1. Open the actor on Apify Console and click Try for free.
2. Paste video or playlist links into Video, playlist, or channel URLs. You can mix videos, playlists and channels in one run.
3. Type a caption language code if you want a specific one (en, es, zh), then Run. Export JSON, CSV or Excel from the dataset.

{
  "video_urls": [
    "https://www.youtube.com/watch?v=arj7oStGLkU",
    "https://youtu.be/UF8uR6Z6KLc",
    "https://www.youtube.com/playlist?list=PLOU2XLYxmsIKpaV8h0AGE05so0fAwwfTw"
  ],
  "language": "en",
  "prefer_manual": true,
  "max_videos_per_source": 50,
  "max_records": 200
}

Minimal input:

{
  "video_urls": ["dQw4w9WgXcQ"]
}

AI agents can call it as the MCP tool herus13--youtube-transcript-scraper on mcp.apify.com, or through the Apify API, with the same JSON. Pass one or more video, playlist or channel links in video_urls, and read transcript_text from each row.

Why this scraper

• Transcript, timestamps and metadata in one row. No chaining a captions tool to a separate metadata scraper and joining the two on a video ID.
• Playlists and channels are valid input. Paste one playlist or channel link and every video in it is found and transcribed, capped where you tell it to stop.
• Ready for a prompt or for a subtitle file. transcript_text is clean newline-separated text; transcript_segments keeps the timings, so .srt and .vtt need no re-timing.
• No API key to obtain. There is no official key that returns caption text, so there is nothing to apply for, meter or rotate.
• You are charged for transcripts, not attempts. A video with no captions returns only a _meta row and is never billed; the run log names it.

Use cases

• RAG pipelines and LLM knowledge bases. A conference playlist or a course becomes a set of transcript_text strings ready to chunk and embed, each keeping the title, channel and publish date a citation needs.
• Podcast and interview transcripts. Hour-long episodes come back as full text with timestamps, ready for show notes, quotes and search.
• Content repurposing. Turn a webinar, lecture or talk into blog drafts, newsletters and social copy by feeding the transcript to a summarizer.
• Video SEO and competitor research. Read what the top videos for a keyword actually say, not what their titles claim, weighted by view_count and published_at.
• Subtitles, translation and accessibility. Convert transcript_segments into .srt or .vtt, or pull an existing Spanish, Chinese or Japanese track after checking available_languages.
• Channel monitoring. Re-run a channel weekly and follow how its talking points change over time.

FAQ

Which languages can I get? Any language the video has a caption track in. Put a language code in Preferred caption language: en, es, fr, de, ja, ko, pt-BR and so on. A short code also matches regional tracks, so zh returns zh-CN or zh-TW. Leave it empty for the video's default track. Every row lists the video's other tracks in available_languages.

Does it translate transcripts? No. It returns caption tracks that exist on YouTube, written by the uploader or generated by YouTube's speech recognition. Many popular videos carry human translations as their own tracks, and those come back like any other language.

Auto-generated or manual captions: which do I get? Human-written captions by default, when the video has them in your language, because they are punctuated and more accurate. Set Prefer human (manual) captions to false to take the auto-generated track. A video with only auto-generated captions returns those whatever the setting says. is_generated on each row tells you which one you got.

What happens with a video that has no captions? It returns only a _meta row with status failed and the reason in message (no captions at all, none in the language you asked for, deleted or private), is named in the run log, and is not charged. A run where the videos simply have no captions ends as succeeded, with no transcript rows; a run where every video was blocked ends as failed.

Can I transcribe a whole playlist or channel? Yes. Paste the playlist link (/playlist?list=…) or channel link (/@handle, /channel/UC…) and its videos are transcribed, newest first for a channel. Max videos per playlist/channel caps each at 100 by default, and Max videos (whole run) caps the run. Mixes and radio lists (list=RD…) and private lists such as Watch Later or Liked are not supported.

Do I get timestamps? Yes. transcript_segments gives every caption line with its start and duration in seconds. transcript_text is the same transcript without timings.

Does it work with YouTube Shorts? Yes. Paste a /shorts/… link or its ID like any other video.

Is a proxy included? A residential proxy is included in the price of the run. To use your own, paste gateway URLs into Your own proxy URLs.

What happens if my input is wrong? The run stops before it scrapes anything, so no results are charged. Its status and log name each field that is wrong and say what to change — fix the input and run again.

Is scraping YouTube transcripts legal? This actor collects only publicly available caption text and metadata that YouTube shows to any viewer. You are responsible for complying with YouTube's terms and applicable law for your use case, including attribution and the copyright of the spoken content itself.

Related actors

Building a content-intelligence pipeline? Pair this actor with:

• Google Trends Scraper — check whether a video's topic is rising before you write about it.
• Google Play App Search & Reviews Scraper — pair video content with what app users actually say.
• Apple App Store Reviews Scraper — the iOS side of the same review picture.
• levels.fyi Salary Scraper — compensation data for career and tech-content research.