YouTube Transcript (RAG-ready)

Get clean transcripts from YouTube videos, playlists and channels — already split by chapter, in Markdown, with token counts. Built for people feeding YouTube into LLMs, RAG pipelines and AI agents.

You only pay for transcripts that are actually returned. Videos without captions, private or deleted videos are reported but never charged.

What you get for each video

Field	Description
status	ok, no_captions, unavailable or error
language, is_generated	Caption language and whether it is auto-generated
segments	[{start, dur, text}] with timestamps in seconds
plain_text	Full transcript as one string
markdown_by_chapter	# Title + ## Chapter sections — ready for chunking
token_count	Tokens (cl100k_base), so you can budget LLM calls
meta	Title, channel, duration, upload date, view count, chapters

Input

• Videos, playlists or channels — any mix of video URLs, 11-character IDs, playlist URLs or channel URLs (https://www.youtube.com/@handle).
• Preferred languages — e.g. ["en", "de"]. Falls back to any available language.
• Max videos per playlist/channel — latest N videos.
• Include metadata and chapters — on by default.
• Bypass cache — results are cached for 7 days; turn this on to always fetch fresh.

{
  "videos": ["https://www.youtube.com/watch?v=kCc8FmEb1nY", "https://www.youtube.com/@3blue1brown"],
  "languages": ["en"],
  "maxPerSource": 20
}

Example output

{
  "video_id": "aircAruvnKk",
  "status": "ok",
  "language": "en",
  "is_generated": false,
  "token_count": 3817,
  "markdown_by_chapter": "# But what is a neural network? | Deep learning chapter 1\n\n## Introduction example\n\n...",
  "meta": {"title": "But what is a neural network? | Deep learning chapter 1", "channel": "3Blue1Brown", "chapters": [{"title": "Introduction example", "start": 0}]}
}

Why this Actor

• Chapter-aware Markdown — markdown_by_chapter follows the video's own chapters, so chunks for RAG keep their meaning.
• Token counts included — budget LLM calls before you send anything.
• No charge for failures — no captions, private or deleted videos are returned with a status and are free.
• Videos, playlists and channels in one input; preferred language with automatic fallback.

Use it from code, n8n or an AI agent

Full runnable examples (LangChain, LlamaIndex, MCP, Python, JS): github.com/1359484419/youtube-transcript-rag-examples

Python

from apify_client import ApifyClient

client = ApifyClient("<YOUR_APIFY_TOKEN>")
run = client.actor("xiaoliang527/youtube-transcript-rag").call(
    run_input={"videos": ["https://www.youtube.com/watch?v=aircAruvnKk"], "languages": ["en"]}
)
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    if item["status"] == "ok":
        print(item["meta"]["title"], item["token_count"])

JavaScript / TypeScript

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor('xiaoliang527/youtube-transcript-rag').call({
  videos: ['https://www.youtube.com/@3blue1brown'], maxPerSource: 10,
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();

LangChain (RAG)

from langchain_apify import ApifyDatasetLoader  # pip install langchain-apify
from langchain_core.documents import Document
from apify_client import ApifyClient

run = ApifyClient("<YOUR_APIFY_TOKEN>").actor("xiaoliang527/youtube-transcript-rag").call(
    run_input={"videos": ["https://www.youtube.com/watch?v=kCc8FmEb1nY"]})
docs = ApifyDatasetLoader(
    dataset_id=run["defaultDatasetId"],
    apify_token="<YOUR_APIFY_TOKEN>",
    dataset_mapping_function=lambda i: Document(
        page_content=i.get("markdown_by_chapter", ""),
        metadata={"video_id": i["video_id"], "title": (i.get("meta") or {}).get("title")}),
).load()
# split by "## " headings and index into your vector store

Claude, Cursor and other MCP clients

Add the Apify MCP server and enable this Actor as a tool:

{
  "mcpServers": {
    "apify": {
      "url": "https://mcp.apify.com?tools=xiaoliang527/youtube-transcript-rag",
      "headers": { "Authorization": "Bearer <YOUR_APIFY_TOKEN>" }
    }
  }
}

Then ask: "Summarize the chapters of https://youtu.be/aircAruvnKk".

n8n / Make / Zapier

Use the official Apify node or app → Run Actor → xiaoliang527/youtube-transcript-rag → Get dataset items.

Pricing

$0.005 per transcript returned ($5 per 1,000). Nothing is charged for videos without captions, private, deleted or invalid videos. Results are cached for 7 days, so re-running the same video returns instantly.

FAQ

Does it work for auto-generated captions? Yes — is_generated tells you which kind you got.

Can I choose the language? Pass languages in order of preference, e.g. ["de", "en"]. If none is available, the first available language is returned.

Shorts and live replays? Supported when they have captions.

Reliability

Every day a regression set of real videos runs against the live Actor; failures by type (blocked, no captions, unavailable) are tracked so breakages are fixed fast.

Limits

Only public captions are returned. The Actor does not download audio or video, does not transcribe audio, and does not access private, members-only or age-restricted content.