YouTube Transcript Extractor & API — bulk channels, playlists & search, RAG-ready chunks

A YouTube transcript extractor you can call as an API: get transcripts for an entire YouTube channel, playlist, or search query in one run — not one video at a time. Every video comes back as a full transcript, timestamped caption segments, and LLM-ready chunks (configurable size and overlap) you can drop straight into a vector database, a RAG pipeline, n8n, or LangChain.

You pay per video and per chunk. Videos with no captions are reported but never charged.

What you get

For each video:

• transcript — the full plain-text transcript
• segments — raw caption segments with start and duration (optional)
• chunks — overlapping text windows with startSec / endSec / approxTokens, ready for embeddings
• metadata: videoId, title, channel, url, language, isGenerated, durationSec, wordCount, chunkCount

Choose one item per video (default), one item per chunk (ideal for vector DBs — every row is a chunk with a deep link …&t=123s), or both.

Input

Field	What it does
startUrls	Any mix of video URLs, channel URLs (https://www.youtube.com/@handle/videos), or playlist URLs
searchQueries	YouTube searches; top results are transcribed
maxVideosPerSource	Cap per channel / playlist / query (default 25)
languages	Preferred language codes in order, e.g. ["en", "es"]. Manual captions are preferred over auto-generated; falls back to whatever exists
outputMode	video, chunks, or both
chunkSizeTokens / chunkOverlapTokens	Chunk geometry (default 500 / 50). Set size to 0 to skip chunking and chunk charges
includeSegments	Include raw timestamped segments in video items
proxyConfiguration	Apify residential proxy recommended for large runs

Example input

{
  "startUrls": [{ "url": "https://www.youtube.com/@lexfridman/videos" }],
  "maxVideosPerSource": 50,
  "languages": ["en"],
  "outputMode": "chunks",
  "chunkSizeTokens": 400,
  "chunkOverlapTokens": 40
}

Output example (one item per video)

{
  "videoId": "dQw4w9WgXcQ",
  "title": "Rick Astley - Never Gonna Give You Up",
  "channel": "Rick Astley",
  "url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
  "language": "en",
  "isGenerated": false,
  "durationSec": 212,
  "wordCount": 371,
  "chunkCount": 1,
  "status": "ok",
  "transcript": "We're no strangers to love ...",
  "chunks": [{ "chunkIndex": 0, "text": "We're no strangers to love ...", "startSec": 18.6, "endSec": 211.9, "approxTokens": 494 }]
}

Pricing

Pay per event, no platform-usage surcharge:

• video — charged once per successfully transcribed video
• chunk — charged per chunk emitted (set chunkSizeTokens to 0 to disable)

A 100-video channel with ~10 chunks per video costs 100 video events + ~1,000 chunk events. Videos without captions cost nothing.

Use with AI agents and automations

• MCP: this Actor is callable as a tool from Claude, ChatGPT, Cursor, and any MCP client through Apify's MCP server.
• n8n / Make / Zapier: use the Apify node and read the dataset.
• API: POST https://api.apify.com/v2/acts/agentbuilt~youtube-transcript-bulk/run-sync-get-dataset-items?token=…

Limits and honesty notes

• Only videos that have captions (manual or auto-generated) can be transcribed. Videos without captions are returned with status: "no_transcript" and are not charged. Audio transcription (Whisper) is planned as an optional add-on.
• YouTube rate-limits aggressive scraping. For runs over ~200 videos, keep the default residential proxy on.
• Member-only, private, or age-restricted videos cannot be transcribed.

About this Actor

Built and maintained by agentbuilt (https://agentbuilt.dev), an AI-operated studio (the code, docs, and support are handled by an AI agent, with a human owner accountable for the account). Report issues in the Issues tab — they are triaged quickly.