YouTube Transcript Scraper

Extract transcripts from any YouTube video via the InnerTube ANDROID client — bypasses YouTube's PoToken wall that broke every web-based transcript scraper in 2026. Plain text, timestamped segments, SRT, VTT, plus full video metadata.

What does YouTube Transcript Scraper do?

Pass video IDs or URLs, get clean transcripts with timestamps in one call. Every record includes:

• Plain transcript text — full video narration as a single string, ready for summarization or RAG ingestion
• Timestamped segments — array of { start, duration, text } for precise navigation
• SRT and VTT output — subtitles-ready formats, selectable per run
• Video metadata — title, author, channel ID, view count, publish date, category, keywords, thumbnail

Why this works when other scrapers fail. Since April 2026, YouTube gates its caption endpoint (/api/timedtext) behind a PoToken — a proof-of-origin token minted by the player's JavaScript. Requests without it return HTTP 200 with an empty body — the worst possible failure mode, because the scraper silently returns nothing. Most libraries broke.

This actor sidesteps the wall entirely. It talks to YouTube's InnerTube ANDROID client (/youtubei/v1/player with clientName: ANDROID) — the same endpoint the Android app uses. YouTube does not enforce PoToken on ANDROID client requests. It's the same approach yt-dlp uses. Runs are backed by residential proxy by default, because YouTube blocks datacenter IPs for caption fetches.

Output fields

Field	Description
videoId	YouTube video ID
videoUrl	Canonical watch URL
title	Video title
author	Channel display name
channelId	Channel ID
lengthSeconds	Video duration in seconds
viewCount	Cumulative views
publishDate	ISO 8601 publication date
category	YouTube category
keywords	Video keywords array
thumbnailUrl	Highest-resolution thumbnail
language	Language code of the transcript
languageName	Human-readable language name
isAutoGenerated	true if ASR auto-captions, false if manual
isTranslatable	Whether the track supports translation
transcript	Plain-text transcript
segments	Array of { start, duration, text }
srt	SRT-formatted subtitles (when outputFormat is srt or all)
vtt	WebVTT-formatted subtitles (when outputFormat is vtt or all)
error	Diagnostic message when a transcript could not be extracted

Who is it for?

• LLM and RAG builders ingesting video content into vector databases
• Content analysts extracting keyword and topic patterns from channel archives
• Podcast and video producers generating subtitles for repurposing
• Market researchers building searchable transcripts across competitor channels
• Academic researchers studying discourse across thousands of videos
• Translation teams pairing transcripts across languages

Pricing

$1.20 per 1,000 results. No subscription.

Results	Cost
100	$0.12
1,000	$1.20
10,000	$12.00

How to use it

1. Paste Video IDs and/or Video URLs.
2. Set Preferred Languages (first match wins).
3. Toggle Include Auto-Generated Captions if manual captions aren't available.
4. Pick Output Format — plain text, SRT, VTT, or all three.
5. Click Start.

Output example

{
  "recordType": "transcript",
  "videoId": "dQw4w9WgXcQ",
  "videoUrl": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
  "title": "Rick Astley - Never Gonna Give You Up (Official Video)",
  "author": "Rick Astley",
  "channelId": "UCuAXFkgsw1L7xaCfnd5JJOw",
  "lengthSeconds": 213,
  "viewCount": 1600000000,
  "publishDate": "2009-10-25T00:00:00-07:00",
  "category": "Music",
  "keywords": ["Rick Astley", "Never Gonna Give You Up", "80s"],
  "thumbnailUrl": "https://i.ytimg.com/vi/dQw4w9WgXcQ/maxresdefault.jpg",
  "language": "en",
  "languageName": "English",
  "isAutoGenerated": false,
  "isTranslatable": true,
  "transcript": "We're no strangers to love. You know the rules and so do I...",
  "segments": [
    { "start": 0.24, "duration": 3.36, "text": "We're no strangers to love" },
    { "start": 3.6, "duration": 3.6, "text": "You know the rules and so do I" }
  ],
  "srt": "1\n00:00:00,240 --> 00:00:03,600\nWe're no strangers to love\n\n2\n...",
  "vtt": "WEBVTT\n\n1\n00:00:00.240 --> 00:00:03.600\n...",
  "error": "",
  "scrapedAt": "2026-09-21T12:00:00.000Z"
}

Technical details

• No API key, no OAuth, no browser. Uses YouTube's public InnerTube endpoints with the ANDROID client context.
• PoToken bypass. The ANDROID client is not subject to PoToken enforcement — this is the core reason the actor works when web-based scrapers return empty bodies.
• Residential proxy by default. YouTube returns "Sign in to confirm you're not a bot" to datacenter IPs for caption fetches. Residential proxy is pre-configured. Disable for one-off tests.
• Caption format negotiation. Requests fmt=json3 (structured JSON) first, falls back to fmt=srv3 (XML) if JSON isn't available.
• Language matching. Manual tracks are preferred over auto-generated tracks. Multiple languages can be ranked by preference.
• Fallback path. When the player endpoint returns no caption tracks, the actor tries the get_transcript continuation endpoint via InnerTube next.

Known limits

• Residential proxy recommended for all but the smallest runs. One-off 5-video tests work on datacenter IPs; anything larger gets rate-limited.
• Some videos have no captions at all. Live streams, some music, and videos the creator disabled captions for return error: "no captions available". These records are still emitted so you have a complete audit trail.
• Auto-generated captions can be imperfect. ASR produces run-on sentences with no punctuation. Manual captions are cleaner when available.
• Translation is not included. The actor returns the track in its native language. If a track is isTranslatable, you can request a translated track via the same InnerTube path — not exposed in this version.
• YouTube changes the API periodically. The InnerTube contract has been stable since 2019. If YouTube updates the client version, patch the ANDROID_CLIENT.clientVersion constant and rebuild.

FAQ

Do I need a YouTube Data API key? No. The Data API doesn't return transcripts anyway — it only returns caption metadata. This actor fetches the actual caption content.

Why do I need a residential proxy? YouTube blocks datacenter IPs from caption endpoints with "Sign in to confirm you're not a bot." Residential proxy is the standard workaround. It's pre-configured; leave it on.

Why did the transcript come back empty? Either the video has no captions, the creator disabled them, or a track in your preferred language doesn't exist. The error field tells you which. Records with empty transcripts still include full metadata.

Can I extract a playlist's transcripts? Yes — resolve the playlist to video IDs first (using a playlist scraper) and pass those IDs as videoIds. Native playlist support is on the roadmap.

How do I get translated captions? Not exposed in this version. The actor returns the native-language track. Translated tracks use a different caption URL parameter — plan to add via the tlang parameter in a future release.

How do I export data? After a run, go to Storage → Export as JSON, CSV, Excel.

Support

Open an issue on the Actor's page for bugs or feature requests.