YouTube Transcript Extractor — No API Key, SRT, Timestamps

Extract YouTube video transcripts as clean, readable text — no duplicated karaoke lines, no timestamp soup, no API key. Paste one URL or a batch, get back prose paragraphs, per-line timestamps, and a ready-to-use SRT string.

Why this one

• Clean text out of the box. YouTube auto-captions are a rolling karaoke buffer: in some caption formats each cue re-emits the tail of the previous one, so naive scrapers return transcripts with phrases duplicated throughout. This actor compares normalized word sequences across cues and removes those overlaps whenever they appear, so what you get reads like prose.
• Original language preferred. If a video was auto-dubbed, asking for "en" on many tools returns a machine translation of the original speech. This actor prefers the track in the language actually spoken (-orig) when one exists.
• Honest billing. You are charged only for videos that actually yield a transcript. Videos with no captions return an error record and cost nothing.
• Fast. No browser, no rendering — a typical video takes well under a second; a 1-hour talk takes a few seconds.

Input

Field	Description
videoUrl	Any YouTube URL (watch, youtu.be, shorts, embed) or a bare 11-character video ID
videoUrls	Batch of videos in one run — one video without captions never stops the rest
language	Preferred caption language code (en, es, de, …), default en
format	paragraphs (readable prose, default) or raw (one block)
includeTimestamps	Adds per-line timestamps and an SRT subtitle string (default true)
maxVideos	Safety cap, 1–200 — also your budget cap since billing is per transcript

Output (one record per video)

{
  "videoId": "dQw4w9WgXcQ",
  "url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
  "ok": true,
  "title": "Rick Astley - Never Gonna Give You Up (Official Video)",
  "channel": "Rick Astley",
  "durationSeconds": 213,
  "language": "en",
  "isAutoGenerated": false,
  "isOriginalLanguageTrack": true,
  "cuesRaw": 61,
  "cuesAfterDedup": 59,
  "duplicationRemovedPct": 3,
  "wordCount": 478,
  "transcript": "We're no strangers to love. You know the rules and so do I...",
  "segments": [{ "start": 18.64, "timestamp": "00:18", "text": "We're no strangers to love" }],
  "srt": "1\n00:00:18,640 --> 00:00:21,880\n..."
}

Videos without captions come back as { "ok": false, "error": "No captions available for this video" } — never charged.

Pricing

$0.004 per transcript. For comparison, the most-used alternatives in this category charge $0.01 (pintostudio/youtube-transcript-scraper) and $0.005 (starvibe/youtube-video-transcript) per transcript (measured 2026-08-07). Videos without captions are never charged.

Common requests this handles

Phrased the way people actually ask, so you can tell before spending anything whether this is the right tool — and so an AI agent picking a tool can match on the real requirement:

• "Get a clean YouTube transcript with no repeated lines." Auto-caption cues overlap; this compares normalized word sequences across cues and drops the duplication, so the text reads as prose paragraphs rather than karaoke.
• "YouTube transcript to SRT subtitles." Every record carries a ready-to-use SRT string alongside the plain text and the per-line timestamps.
• "Transcript with timestamps for each line." Per-cue start/duration are returned as structured fields, not baked into the text.
• "Transcript in the language actually spoken, not a dub." When a video was auto-dubbed the original-language track (-orig) is preferred over a machine translation.
• "Bulk YouTube transcripts for a list of video URLs." Pass a batch; each video becomes its own record, and videos with captions disabled are reported and never charged.
• "YouTube captions for a RAG pipeline / to feed an LLM." Plain readable text is the default output, which is what a model wants — you are not paying tokens for duplicated phrases.
• "Transcript without an API key or cookies." No YouTube Data API quota, no login, no cookies.

Common uses

• Feed video content to an LLM (summaries, Q&A, RAG pipelines) without paying for duplicated text
• Subtitle files (SRT) for re-uploads, translations, or accessibility
• Research and content analysis across channels
• Show notes, blog posts, and quotes from talks and podcasts

Notes

• Works with auto-generated (ASR) and manually uploaded captions, in any language YouTube serves.
• Extraction talks to YouTube's own player API — no rendering, no third-party transcript services.
• If YouTube has no captions at all for a video (some music videos, very new uploads), there is nothing to extract; you are not charged for those.

Use from code or AI agents

Run it from any HTTP client and get the transcripts back in the same call:

curl -X POST "https://api.apify.com/v2/acts/eliai~youtube-transcript-clean/run-sync-get-dataset-items?token=$APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"videoUrl": "https://www.youtube.com/watch?v=dQw4w9WgXcQ", "format": "paragraphs"}'

Or with the Apify JS client:

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor('eliai/youtube-transcript-clean').call({
  videoUrls: ['dQw4w9WgXcQ', 'jNQXAC9IVRw'],
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();

AI agents can call this actor as a tool through Apify's MCP server — an MCP-enabled agent can discover it, read the input schema, and fetch transcripts autonomously.

FAQ

How do I get the transcript of a YouTube video as plain text? Paste the video URL into videoUrl and run — the transcript field of the result is clean prose with duplicated caption fragments removed.

Does it work on videos without manual subtitles? Yes — auto-generated (ASR) captions are supported in any language YouTube serves. If a video has no captions at all, you get an ok: false record and are not charged.

Can I download subtitles as an SRT or VTT file? Yes. With includeTimestamps on (the default), every result includes srt and vtt strings — save them as subtitles.srt or subtitles.vtt.

Does it work with YouTube Shorts and youtu.be links? Yes — watch, youtu.be, shorts, and embed URLs all work, as do bare 11-character video IDs.

Why do transcripts from other tools repeat phrases twice? YouTube auto-captions are a rolling karaoke buffer; in some formats each cue re-emits the tail of the previous one. This actor compares normalized word sequences across cues and removes those overlaps, so the text reads like prose.

How do I transcribe a whole list of videos at once? Put up to 200 URLs in videoUrls — one captionless video never stops the rest, and you pay only per transcript actually extracted.

Can I get the transcript in a specific language? Set language (e.g. es, de). If a video was auto-dubbed, the track in the language actually spoken is preferred over a machine translation of it.

For AI agents

This Actor is built to be called by software, not just by people.

• Mount it directly as an MCP tool — no Store search, no ranking, just this one tool: https://mcp.apify.com/?actors=eliai/youtube-transcript-clean
• Or call it over HTTP and get the results in the same request: POST https://api.apify.com/v2/acts/eliai~youtube-transcript-clean/run-sync-get-dataset-items
• Pay with x402, without an Apify account. This Actor is whitelisted for agentic payments, so an agent holding USDC on Base can buy a prepaid token and spend it here. The minimum purchase is $1, the token balance is an absolute spending cap, and it expires 14 days after purchase.
• Costs are predictable before you call. Pricing is pay-per-event (see Pricing above), so an agent can budget a run in advance instead of discovering the bill afterwards.
• Send only the field you mean. If you pass the bulk field, it is used on its own; the single-value field is a fallback, never merged into your request. You are charged for the items you sent and nothing else.

Changelog

• 2026-08-27 — Added WebVTT (vtt) next to SRT. Added the TV embedded InnerTube client as a third caption path so a mobile-client block no longer jumps straight to residential proxy. Runs now stop cleanly on the platform deadline instead of timing out.