📝 YouTube Transcript Scraper: Captions to Clean JSON (No API Key)

Part of the Social & Market Research MCP. This actor's data is also available to AI agents through our Social & Market Research MCP server — eight social, news and search-interest tools behind one endpoint. No result, no charge.

⚡ 16/16 runs succeeded in the last 30 days · no API key, no OAuth, no quota.

💸 You're only charged for delivered results. Empty searches, failed pages and duplicate listings are never billed.

Overview

YouTube Transcript Scraper pulls timestamped transcripts and captions from any public YouTube video and returns them as clean, structured JSON, ready to drop straight into a RAG pipeline, a vector store, or an LLM prompt. Give it a list of video URLs (or bare IDs) and get back per-video segments, a joined fullText, the caption language, and a character count.

No API key, no OAuth, no quota. It works on any public video, including auto-generated captions.

✅ No login required | ✅ No API key | ✅ Pay only for delivered transcripts | ✅ MCP-ready for AI agents

Features

Structured JSON output. Per-segment { start, dur, text } plus a joined fullText. Any URL shape. watch?v=, youtu.be, /shorts/, /embed/, or bare 11-char IDs. Language preference with fallback. Prefer en, es, hi and fall back to the default track. Auto vs. human captions flagged. isAutoGenerated tells you which you got. Free failure handling. Videos with captions disabled return no-captions and are never billed.

How it works

The official YouTube Data API caption endpoints require OAuth, channel ownership, and a daily quota. You effectively cannot download the caption text of videos you do not own. This scraper reads the same public caption tracks YouTube already serves to any viewer's player. No key, no OAuth, no quota.

For each video, the actor resolves the canonical ID, reads available caption tracks from the watch page, and pulls the timed-text XML for the preferred language (falling back if that language is missing). Segments are cleaned into { start, dur, text } items in seconds and joined into a readable fullText. Videos with captions disabled return status: no-captions and are not billed.

🧾 Input configuration

{
  "videoUrls": [
    "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
    "https://youtu.be/9bZkp7q19f0",
    "kJQP7kiw5Fk"
  ],
  "language": "en",
  "includeTimestamps": true,
  "proxy": { "useApifyProxy": true }
}

📤 Output format

{
  "videoId": "dQw4w9WgXcQ",
  "url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
  "title": "Rick Astley, Never Gonna Give You Up (Official Video)",
  "language": "en",
  "isAutoGenerated": false,
  "segments": [
    { "start": 18.8, "dur": 3.2, "text": "We're no strangers to love" },
    { "start": 22.0, "dur": 3.36, "text": "You know the rules and so do I" }
  ],
  "fullText": "We're no strangers to love You know the rules and so do I ...",
  "charCount": 1542,
  "segmentCount": 84,
  "status": "ok",
  "scrapedAt": "2026-07-10T09:15:00.000Z"
}

Every transcript record contains these fields:

Field	Description
🆔 videoId	11-character YouTube video ID
🔗 url	Canonical watch URL
🏷️ title	Video title (null if unparsable)
🌐 language	Language code of the caption track used
🤖 isAutoGenerated	True for auto (ASR) captions, false for human/uploaded
⏱️ segments	Ordered { start, dur, text } items in seconds
📝 fullText	All segment text joined into one readable string
🔢 charCount	Character length of fullText
📊 segmentCount	Number of transcript segments
🚦 status	ok, no-captions, or error
🕒 scrapedAt	ISO 8601 timestamp of the fetch

The run also pushes a final status: "summary" record with counts (transcriptsScraped, noCaptions, errored, chargedFor).

💼 Common use cases

RAG and vector search Chunk fullText, embed it, and let an LLM answer questions grounded in video content. Chain into the rag-crawler to index entire channels.

Video summarization Pipe transcripts to an LLM for TL;DRs, chapter markers, or highlight reels. Batch process a channel to produce weekly digests.

Content repurposing Turn webinars, podcasts, and tutorials into articles, show notes, and social posts. Feed the segments[] array into a subtitle or translation workflow.

Research and dataset building Mine spoken content across many videos for topics, keywords, and tone. Assemble timestamped speech-to-text corpora for fine-tuning or analysis.

🚀 Getting started

1. Open the actor in Apify Console (or call it via API or MCP).
2. Under YouTube video URLs or IDs, paste one or more videos: watch URLs, youtu.be links, /shorts/, /embed/, or bare IDs.
3. Set preferred caption language (e.g. en, es, hi). The actor falls back to the default track if that language is missing.
4. Toggle Include timestamps on for { start, dur, text } segments, or off for fullText only.
5. Click Save and Start, then download the dataset as JSON, CSV, or Excel, or pull via API or MCP.

💵 Pricing

One pay-per-event charge: transcript-scraped — charged when a transcript comes back with status: "ok".

Apify applies a discount tier to your account. Price per transcript at each tier:

Apify tier	Per transcript	Per 1,000 transcripts
FREE	$0.001	$1.00
BRONZE	$0.00084	$0.84
SILVER	$0.00071	$0.71
GOLD	$0.0006	$0.60
PLATINUM	$0.0006	$0.60
DIAMOND	$0.0006	$0.60

At the FREE tier that is $1.00 per 1,000 transcripts. Your tier is shown on your Apify billing page.

Pay-per-event, no subscription and no monthly minimum. Videos with captions disabled (status: "no-captions"), errored videos, and failed or empty runs are never billed — you only pay for a transcript actually returned.

⏰ Run it on a schedule

Transcribing a channel's new uploads as they publish is the natural recurring use — the transcripts land in one growing dataset ready for a RAG index or a weekly digest.

1. On this actor's page, click ⋯ → Schedule actor (or Console → Schedules → Create new).
2. Pick a frequency — @weekly fits most channel digests; @daily for high-volume channels.
3. Your saved input is reused on every run; each run appends to a named dataset.
4. Wire the dataset to Google Sheets, Slack, or a webhook via the actor's Integrations tab so new records reach you automatically.

FAQ

Do I need a YouTube API key or account? No. The scraper reads public caption tracks directly from the watch page and the public timedtext endpoint. No API key, no OAuth, no login, and no quota.

What video URL formats are supported? Full watch?v= URLs, youtu.be/… short links, /shorts/…, /embed/…, and bare 11-character video IDs. Each resolves to the canonical video automatically.

What happens if a video has no captions? The record comes back with status: "no-captions" and is not charged. Only videos that return an actual transcript are billed.

Can I choose the caption language? Yes. Set preferred caption language to a two-letter code (e.g. en, es, hi, fr). The actor prefers an exact match, then a language-prefix match (en matches en-US), then the video's default track, then the first available, and reports what it used in language.

Are auto-generated (ASR) captions supported? Yes. When only auto captions exist, the actor returns them and sets isAutoGenerated: true, so you can tell human captions apart from machine ones.

How is it priced? Pay per result: one charge per transcript actually returned. There is no free tier. Videos with captions disabled, and failed or empty runs, are never charged.

Can I use it inside an AI agent? Yes. It is exposed as an MCP tool. See below.

How do I get YouTube transcripts in bulk without the API? Paste every video into videoUrls in a single run — there is no per-video copy-and-paste and no OAuth handshake. The actor reads the same public caption tracks YouTube serves to any viewer, so a list of hundreds of videos becomes one dataset you can export as JSON, CSV, or Excel, or pull straight into an agent over MCP.

Can I download the transcript of a video I don't own? Yes, for any public video that has captions. This is the main gap this actor fills: YouTube's official Data API caption-download endpoints are scoped to channels you own, so they cannot return the caption text of someone else's video. Reading the public caption track has no such restriction.

Can I get the transcript with timestamps? Yes. Leave includeTimestamps on and every record carries a segments array of { start, dur, text } items in seconds, alongside the joined fullText. Turn it off if you only want the plain text block.

Use in Claude, ChatGPT & any MCP agent

https://mcp.apify.com/?tools=themineworks/youtube-transcript-scraper

Or call it programmatically with the Apify client:

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: 'YOUR_APIFY_TOKEN' });

const run = await client.actor('themineworks/youtube-transcript-scraper').call({
  videoUrls: ['https://www.youtube.com/watch?v=dQw4w9WgXcQ'],
  language: 'en',
  includeTimestamps: true,
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items);

🛠️ Complete your YouTube pipeline

Pair the transcript scraper with the rest of the video suite:

• YouTube Channel Scraper: subscribers, video list, and channel stats, no API key.
• RAG Crawler: index entire sites for LLM retrieval.
• Reddit Scraper: pull public posts and comment trees for training data.

Typical flow: youtube-channel discovers the videos, youtube-transcript-scraper turns them into text, the RAG crawler assembles the wider corpus.

Found a bug or have a feature request? Open an issue on the actor's Apify Console page or reach out through the Apify profile.