YouTube Subtitle Scraper — Transcripts & Captions

Thirdwatch portfolio: 5K users across 97 public Actors, 2M+ records delivered, and >99% run success. Explore all Thirdwatch Actors.

Extract subtitles and transcripts from public YouTube videos as clean, timestamped segments. Works with both manually uploaded captions and YouTube's auto-generated tracks, in any language the video offers — no YouTube API key required.

Use it for transcript datasets, video summarization, translation pipelines, accessibility audits, and content research.

Why this Actor

• Manual + auto-generated captions: prefers human-written tracks when preferManual is set, falls back to auto-generated.
• Any language: pass ISO codes like en, hi, es — or omit to take the best available track and get the full availableLanguages list in output.
• Timestamped segments: every row carries the full segment list (startMs, durationMs, text) plus a joined transcript string.
• Per-video sticky sessions: each video gets its own residential proxy session with a fresh-exit retry on rate limits, so multi-video batches stay reliable.
• Rich video metadata: title, channel, channelId, duration, view count, and upload date on every row.

Quick start

{
  "videoUrls": ["https://www.youtube.com/watch?v=dQw4w9WgXcQ"],
  "languages": ["en"],
  "preferManual": true,
  "proxyCountry": "US"
}

Returns one dataset row per video with all caption segments.

Input

Field	Required	Description
videoUrls	Yes	YouTube watch URLs or video IDs.
languages	No	Preferred caption languages, ISO 639-1 (e.g. en, hi). Default: first available.
preferManual	No	Prefer human-uploaded captions over auto-generated. Default: true.
proxyCountry	No	Two-letter residential proxy country. Recommended for batches — YouTube rate-limits shared exits.

Output

Field	Description
videoId, url	Video identity and canonical URL.
title, channel, channelId	Video and channel metadata.
duration, viewCount, uploadDate	Length in seconds, views, publish date.
language, captionSource	Track language and whether it is manual or auto.
availableLanguages	All caption languages the video offers.
segments	List of {startMs, durationMs, text} caption segments.
segmentCount, transcript	Segment count and the joined plain-text transcript.

Example row:

{
  "videoId": "dQw4w9WgXcQ",
  "url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
  "title": "Rick Astley - Never Gonna Give You Up (Official Video)",
  "channel": "Rick Astley",
  "language": "en",
  "captionSource": "manual",
  "segmentCount": 61,
  "segments": [{"startMs": 18170, "durationMs": 2290, "text": "We're no strangers to love"}],
  "transcript": "We're no strangers to love ..."
}

Notes & limitations

• Heavily scraped videos occasionally hit YouTube 429 on the caption file itself; the Actor retries once on a fresh residential exit. A small fraction of videos may still fail — they return an error row rather than failing the run.
• Residential proxy traffic is the main cost; expect ~$0.01 per video.