YouTube Transcript Scraper - Subtitles, Chapters & AI/RAG API
Pricing
from $8.00 / 1,000 extracted youtube video transcript & ai/rag chunks
YouTube Transcript Scraper - Subtitles, Chapters & AI/RAG API
Fast YouTube Transcript Scraper & Subtitle Extractor API. Extract full text, timestamps, video chapters, metadata & semantic RAG chunks. 3-tier self-healing failover. No login needed; works with Apify MCP, LangChain, n8n & Make.
Pricing
from $8.00 / 1,000 extracted youtube video transcript & ai/rag chunks
Rating
0.0
(0)
Developer
H H
Maintained by CommunityActor stats
1
Bookmarked
1
Total users
0
Monthly active users
7 days ago
Last modified
Categories
Features
Share
What can YouTube Transcript Scraper do?
YouTube Transcript Scraper extracts clean spoken transcripts, word-accurate timestamps, video chapters, channel metadata, and token-bounded AI/RAG chunks from any public YouTube video or Short in a single run. Powered by a 3-tier self-healing extraction cascade (youtube-transcript-api -> Android InnerTube JSON3 -> yt-dlp subtitle tracks) inside a lightweight 256 MB container, it delivers reliable subtitle extraction with zero browser overhead:
| 📦 What you can extract | ⚙️ Features & integrations |
|---|---|
|
📝 Full clean transcripts with rolling-caption deduplication ( |
🔄 3-tier self-healing engine that automatically fails over across 3 caption endpoints |
Whether you are indexing thousands of podcasts and technical talks into a RAG vector database, summarizing competitor product launches, repurposing long-form YouTube videos into newsletters and social threads, or building timestamp-cited AI research agents, YouTube Transcript Scraper gives you clean, deduplicated text in seconds.
What data can I extract with YouTube Transcript Scraper?
YouTube Transcript Scraper pulls structured public subtitle tracks and video context without requiring a Google Cloud OAuth token, YouTube Data API v3 quota, or logged-in cookies. Every extracted video item gives your pipeline both human-readable full text and machine-ready timestamped chunks:
-
Clean full-text transcript: deduplicated spoken text ready for summarization. Raw YouTube auto-captions repeat overlapping phrases across consecutive cue windows. Our cleaner strips duplicate rolling prefixes/suffixes and HTML entities into
fullText, alongsidewordCount,characterCount,language,isAutoGenerated, andengineUsed. -
Semantic AI / RAG chunks: citation-ready blocks for vector embeddings. Each item in
ragChunksbundles consecutive subtitle cues up tochunkSizeWords(default200words) and attacheschunkIndex,text,wordCount,startTime,endTime, and a clickabledeepLinkUrl(https://youtu.be/<videoId>?t=120s) so your LLM answers cite the exact second in the video. -
Fine-grained subtitle segments: utterance-level timing for video clipping and search. Both
segmentsandtranscript(included as an alias for drop-in workflow compatibility) provide{text, start, duration}arrays measured in floating-point seconds. -
Creator-defined chapters: structural outline of long-form videos. When
includeMetadatais enabled and the creator added chapter markers in the description,chaptersreturns{title, startTime, endTime, deepLinkUrl}for every section. -
Video & channel context: who published it and how it performed. Extracts
videoId, canonicalvideoUrl,title,channelName,channelId,publishDate(YYYYMMDD),viewCount,likeCount, anddurationSeconds.
Need more specialized workflows? Combine this Actor with these dedicated tools from our suite:
- 📧 YouTube Channel Email Scraper: extract verified business emails, sponsor links, Linktrees & social handles from YouTube channels
- 🕸️ Fast Website Content Crawler: convert any documentation site or blog into matching LLM-ready Markdown & RAG chunks
- 📰 Google News Scraper Fast: decode Google News publisher URLs and extract full-article AI/RAG chunks
- 🎵 TikTok Viral Hook & Script Scraper: dissect short-form TikTok WebVTT spoken hooks, body scripts & CTAs
How much will YouTube Transcript Scraper cost?
YouTube Transcript Scraper uses a transparent Pay-Per-Event (PPE) pricing model, where you pay a tiny $0.05 Actor Start event plus a per-item fee for each Extracted YouTube Video Transcript & AI/RAG Chunks. On the Apify Free plan ($5.00 monthly free credit included), it costs $0.025 per item ($25.00 per 1,000 results) — letting you extract up to 198 results every month for free.
Paid Apify plans automatically unlock volume discounts up to 68% off ($0.012 on Starter/Bronze, $0.01 on Scale/Silver, and $0.008 on Business/Gold, Platinum & Diamond) with zero residential proxy surcharges:
| Apify Plan / Tier | Price per extracted dataset item | Price per 1,000 results | Batch Depth & Unmasking | Approx. Monthly Capacity |
|---|---|---|---|---|
Free (FREE Trial Preview) | $0.025 | $25.00 | 5 items/run trial (items 1-2 100% unmasked, items 3-5 preview-masked) | ~198 trial results |
Starter (BRONZE — $29/mo) | $0.012 | $12.00 | Unlimited batch runs & 100% unmasked | ~2,412 results |
Scale (SILVER — $199/mo) | $0.01 | $10.00 | Unlimited batch runs & 100% unmasked | ~19,895 results |
Business (GOLD — $999/mo) | $0.008 | $8.00 | Unlimited batch runs & 100% unmasked | ~124,868 results |
Enterprise (PLATINUM / DIAMOND) | $0.008 / $0.008 | $8.00 | Unlimited batch runs & 100% unmasked | Custom high-volume SLA |
Free Trial vs. Paid Plan Unlock: On the Apify Free tier, runs execute in 5-item Trial Preview mode (items
1and2are returned100%unmasked so you can verify live email/data quality firsthand; items3to5mask emails, phones, and deep RAG/financial blobs). Upgrading to Apify Starter ($29/mo) or higher immediately unlocks unlimited batch extraction (1,000+items/run),100%unmasked emails and payloads, and up to 68% lower per-item pricing ($0.012on Starter down to$0.008on Business). Check the pricing tab for full details.
How do I use YouTube Transcript Scraper?
YouTube Transcript Scraper was designed to be effortless to start with — even if you have never configured proxies, cookies, or web scrapers before. Here is how to extract data in under a minute:
- Create a free Apify account using your email (includes
$5.00in monthly free usage credit). - Open YouTube Transcript Scraper in Apify Console.
- Configure your target inputs (
videoUrl,videoUrls,targetLanguage) in the Input tab (use Bulk edit to paste lists at once). - Adjust your results limit or optional enrichment filters to match your workflow.
- No proxy setup is required — the Actor runs directly in a fast 256 MB container out of the box.
- Click Save & Start and watch structured records stream into your Dataset in real time.
- Open the Storage tab to inspect your results as a clean table or raw JSON.
- Download your dataset in JSON, CSV, Excel (
.xlsx), or XML, or connect it directly to Clay, n8n, Make, Zapier, Google Sheets, or MCP.
⬇️ Input
YouTube Transcript Scraper accepts a single video (videoUrl), a batch list of videos (videoUrls), or both in the same run (all inputs are merged and deduplicated automatically by 11-character YouTube Video ID before extraction).
Mode 1: Single Video Drop-In (videoUrl)
Ideal for n8n, Make, Zapier, or MCP tool calls triggered whenever a new YouTube URL is posted.
| Input / Configuration | What it returns |
|---|---|
Standard watch URL (https://www.youtube.com/watch?v=IELMSD2kdmk) | Full transcript, metadata, chapters & RAG chunks |
Shortened share link (https://youtu.be/IELMSD2kdmk) | Normalized to 11-char videoId and extracted |
YouTube Shorts URL (https://www.youtube.com/shorts/...) | Spoken Short caption track & metadata |
Bare 11-character Video ID (IELMSD2kdmk) | Resolved directly without extra redirects |
Input JSON example:
{"videoUrl": "https://www.youtube.com/watch?v=IELMSD2kdmk","targetLanguage": "en","chunkSizeWords": 200,"includeMetadata": true}
Mode 2: High-Speed Batch Extraction (videoUrls)
Pass dozens or hundreds of YouTube URLs or IDs in one run; up to 6 videos are processed concurrently in worker threads.
| Input / Configuration | What it returns |
|---|---|
Array of URLs (["https://www.youtube.com/watch?v=...", ...]) | One structured dataset record per unique video |
Target language ISO code ("es", "de", "fr", "pt") | Selects native subtitle track or auto-translated track |
Input JSON example:
{"videoUrls": ["https://www.youtube.com/watch?v=IELMSD2kdmk","https://www.youtube.com/watch?v=2DXQqwKSwJE"],"targetLanguage": "en","chunkSizeWords": 250,"includeMetadata": true}
Complete Input Parameter Reference
| Parameter | Type | Default | Required | Description |
|---|---|---|---|---|
videoUrl | string | "https://www.youtube.com/watch?v=2DXQqwKSwJE" | No | A single YouTube video URL, Short URL, or 11-character Video ID (100% compatible with existing workflows). |
videoUrls | array | [] | No | Pass multiple YouTube video URLs or IDs in a single run for high-speed parallel extraction. |
targetLanguage | string | "en" | No | Preferred transcript language code (e.g., 'en', 'es', 'fr', 'de', 'pt'). Automatically translates if native track is in another language. |
chunkSizeWords | integer | 200 | No | Number of words per semantic RAG chunk (with exact timestamp deep-links). Set to 150-300 for optimal vector database embedding. |
includeMetadata | boolean | true | No | Extract video title, channel info, views, publish date, description, and timestamped chapters. |
⬆️ Output and setup
All extracted records stream incrementally into your run's Dataset under the Storage tab in Apify Console. Preview results as a clean interactive table, download in JSON, JSONL, CSV, Excel (.xlsx), XML, or RSS, or consume items directly via the Apify REST API, Python/Node.js SDKs, or MCP.
📦 Extracted YouTube video transcript + AI/RAG chunks (status: "success")
{"status": "success","billable": true,"videoId": "IELMSD2kdmk","videoUrl": "https://www.youtube.com/watch?v=IELMSD2kdmk","title": "Apache Spark in 100 Seconds","channelName": "Fireship","channelId": "UCsBjURrPoezykLs9EqgamOA","publishDate": "20241119","viewCount": 600334,"likeCount": 20985,"durationSeconds": 199,"language": "en","isAutoGenerated": true,"engineUsed": "engine1_youtube_transcript_api","wordCount": 673,"characterCount": 3826,"chunkCount": 4,"fullText": "Apache spark an open- Source data analytics engine that can process massive streams of data from multiple sources like an octopus juggling chainsaws it was created in 2009 by mate zaharia at UC Berkeley's amp lab around this time the amount of data being collected on the internet was exploding from megabytes to pedabytes making it impossible to analyze on...","chapters": [{"title": "Introduction to Spark","startTime": 0.0,"endTime": 37.0,"deepLinkUrl": "https://youtu.be/IELMSD2kdmk?t=0s"},{"title": "In-memory processing power","startTime": 37.0,"endTime": 60.0,"deepLinkUrl": "https://youtu.be/IELMSD2kdmk?t=37s"}],"ragChunks": [{"chunkIndex": 0,"startTime": 0.32,"endTime": 63.28,"timestampRange": "00:00 - 01:03","deepLinkUrl": "https://youtu.be/IELMSD2kdmk?t=0s","wordCount": 200,"text": "Apache spark an open- Source data analytics engine that can process massive streams of data from multiple sources like an octopus juggling chainsaws it was created in 2009 by mate zaharia at UC Berkeley's amp lab around this time the amount of data being collected on the internet was exploding from megabytes to pedabytes making it impossible to analyze on..."},{"chunkIndex": 1,"startTime": 61.52,"endTime": 122.0,"timestampRange": "01:01 - 02:02","deepLinkUrl": "https://youtu.be/IELMSD2kdmk?t=61s","wordCount": 206,"text": "spark locally on your own machine it's written in Java and runs on the jvm but its apis can be used with wrappers for python SQL and many other languages to get started install it then let's imagine we have a CSV file with four columns for City population latitude and longitude and our boss wants us to find the city with the biggest population between the..."}],"segments": [{"start": 0.32,"duration": 4.08,"end": 4.4,"timestamp": "00:00","text": "Apache spark an open- Source data"},{"start": 2.48,"duration": 3.84,"end": 6.32,"timestamp": "00:02","text": "analytics engine that can process"}],"transcript": [{"start": 0.32,"duration": 4.08,"end": 4.4,"timestamp": "00:00","text": "Apache spark an open- Source data"},{"start": 2.48,"duration": 3.84,"end": 6.32,"timestamp": "00:02","text": "analytics engine that can process"}]}
📦 Zero-charge captionless video guardrail (billable: false)
{"status": "no_transcript_available","videoId": "M7lc1UVf-VE","videoUrl": "https://www.youtube.com/watch?v=M7lc1UVf-VE","title": "YouTube Developers Live: Embedded Web Player Customization","channelName": "Google for Developers","error": "No captions or transcripts exist on this video.","billable": false}
Output Field Dictionary
| Output Field | Type | Description |
|---|---|---|
status | string | "success", "no_transcript_available", or "error". |
billable | boolean | true only when spoken transcript words were extracted; false on captionless/invalid videos. |
videoId | string | Canonical 11-character YouTube Video ID. |
videoUrl | string | Normalized https://www.youtube.com/watch?v=<videoId> URL. |
title | string | Video title (via InnerTube / yt-dlp or YouTube oEmbed fallback). |
channelName | string | Creator or brand channel display name. |
channelId | string | YouTube UC... channel identifier. |
publishDate | string | Upload date in YYYYMMDD format when includeMetadata is enabled. |
viewCount | integer | Public view count at the moment of extraction. |
likeCount | integer | Public like count at the moment of extraction. |
durationSeconds | integer | Total video runtime in seconds. |
language | string | Resolved subtitle track language code (e.g., "en"). |
isAutoGenerated | boolean | true if extracted from YouTube ASR automatic captions; false if human-uploaded. |
engineUsed | string | Which tier succeeded (engine1_youtube_transcript_api, engine2_android_innertube, or engine3_ytdlp_json3). |
wordCount | integer | Total word count of deduplicated fullText. |
characterCount | integer | Total character length of fullText. |
chunkCount | integer | Number of semantic chunks in ragChunks. |
fullText | string | Complete, deduplicated plain-text transcript. |
chapters | array | Array of {title, startTime, endTime, deepLinkUrl} video chapters. |
ragChunks | array | Array of {chunkIndex, text, wordCount, startTime, endTime, deepLinkUrl} objects. |
segments / transcript | array | Utterance-level {text, start, duration} timing segments. |
Notes, tips, and exceptions
- Zero-charge guardrail on silent/captionless videos: If a video is music-only or the uploader disabled all subtitles, the Actor emits
status: "no_transcript_available"withbillable: falseand does not trigger the dataset-item charge event. - Automatic URL deduplication: Passing both
videoUrlandvideoUrls(or duplicate links invideoUrls) automatically deduplicates inputs so you never pay twice for the same video. - Rolling-caption cleanup: Auto-generated YouTube captions often emit two-line rolling cues where line 1 repeats line 2 of the previous event. Our segment cleaner removes overlapping runs before building
fullTextandragChunks. - Faster metadata-free mode: Setting
includeMetadata: falseskips full metadata/chapter enrichment when an early transcript engine (engine1orengine2) succeeds, cutting per-video latency even further. - Language fallback order: If your requested
targetLanguageis unavailable natively, the extractor checks YouTube's automatic translation track, then falls back cleanly to English (en,en-US,en-GB) or the primary native track. - Embeds, Shorts & live replays: Standard watch URLs,
youtu.belinks,/shorts/,/embed/, and/live/VOD replays with archived captions are all supported. - Private, members-only, or age-gated videos: Only publicly accessible videos can be extracted without cookies.
- Snapshot metrics:
viewCountandlikeCountreflect public YouTube counters at the timestamp of the run.
What can you do with data from YouTube Transcript Scraper?
YouTube Transcript Scraper powers high-ROI automation, lead generation, and AI agent pipelines. Here are four proven step-by-step recipes:
🧠 Build a timestamp-cited YouTube RAG knowledge base
Turn a playlist, technical conference channel, or podcast archive into an AI assistant that answers questions with clickable timestamp citations.
Recipe:
-
Collect target video URLs. Export video links from a YouTube playlist, RSS feed (
youtube.com/feeds/videos.xml?channel_id=...), or search workflow. -
Extract transcripts with 200-word RAG chunks. Run YouTube Transcript Scraper with
videoUrls,chunkSizeWords: 200, andincludeMetadata: true. -
Embed
ragChunksinto your vector store. Iterate over each item'sragChunks[], embedchunk.textwith OpenAI/Voyage/Cohere, and storevideoId,title,channelName, andchunk.deepLinkUrlas metadata in Pinecone, Qdrant, Supabasepgvector, or Weaviate. -
Serve cited answers. When your LLM synthesizes an answer, surface
deepLinkUrl(https://youtu.be/...?t=120s) so users jump straight to the exact spoken moment.
Schedule daily via Apify Cron or n8n to automatically ingest newly uploaded videos from tracked channels.
✍️ Repurpose long-form podcasts into newsletters, LinkedIn posts & blog articles
Content agencies and B2B founders use spoken transcripts + chapter boundaries to turn one 45-minute video into a week of written assets.
Recipe:
-
Pull chapters and clean full text. Run the Actor on your latest episode URL with
includeMetadata: true. -
Segment by
chapters[]orragChunks[]. Feed each chapter range alongsidefullTextinto Claude or GPT-4o inside Make.com or n8n. -
Generate multi-channel drafts. Prompt your LLM to extract quotable hooks, a structured Substack issue, and 5 short social posts anchored to real spoken quotes.
Deliver drafts straight to Notion, Google Docs, or Slack for 5-minute human editing.
🔎 Competitor product demo & webinar intelligence
Competitors often reveal roadmap features, pricing objections, and customer case studies inside webinars and YouTube tutorials long before updating their marketing site.
Recipe:
-
Track competitor uploads. Monitor competitor YouTube channels weekly.
-
Batch-extract spoken transcripts. Pass new uploads in
videoUrlsand filterstatus == "success". -
Run automated feature & pricing extraction. Have an AI agent scan
fullTextfor feature launches, partner names, and positioning shifts.
Route weekly digests into your product marketing Slack channel or competitive battlecards.
🤝 Vet & pitch YouTube creators with hyper-personalized sponsorship hooks
Cold emails that reference something specific a creator actually said in their latest video convert dramatically higher than generic templates.
Recipe:
-
Discover niche channels & verified emails. Run YouTube Channel Email Scraper on your niche keyword (e.g.,
"AI automation agency") to getprimaryEmailandrecentVideosSample. -
Extract the creator's latest video transcript. Pass the top video URL from each channel into YouTube Transcript Scraper.
-
Generate personalized 1-sentence icebreakers. Feed the first 2
ragChunksinto Clay or n8n to reference the creator's exact thesis in your sponsorship pitch.
Sync enriched leads directly to Instantly, Smartlead, HubSpot, or Lemlist.
Want to get other B2B, Social, Financial, or MCP Agent data?
Chain this tool with our other 128 MB / 256 MB zero-proxy Apify Scrapers and Live Streamable-HTTP MCP Servers:
🔌 Using YouTube Transcript Scraper with the Apify API, SDKs & MCP
Automate YouTube Transcript Scraper programmatically from Python, Node.js/TypeScript, cURL, workflow builders (n8n, Make, Zapier, Clay, Pipedream), or Model Context Protocol (MCP) AI clients.
Python (apify-client)
from apify_client import ApifyClientclient = ApifyClient("<YOUR_APIFY_API_TOKEN>")run = client.actor("church_of_loujess/youtube-transcript-rag-scraper").call(run_input={"videoUrl": "https://www.youtube.com/watch?v=IELMSD2kdmk","targetLanguage": "en","chunkSizeWords": 200,"includeMetadata": True})for item in client.dataset(run["defaultDatasetId"]).iterate_items():print(item)
Node.js / TypeScript (apify-client)
import { ApifyClient } from 'apify-client';const client = new ApifyClient({ token: '<YOUR_APIFY_API_TOKEN>' });const run = await client.actor('church_of_loujess/youtube-transcript-rag-scraper').call({"videoUrl": "https://www.youtube.com/watch?v=IELMSD2kdmk","targetLanguage": "en","chunkSizeWords": 200,"includeMetadata": true});const { items } = await client.dataset(run.defaultDatasetId).listItems();console.log(`Extracted ${items.length} items:`, items);
One-Line Synchronous REST API (cURL — ideal for n8n, Make & Clay HTTP Columns)
curl -X POST "https://api.apify.com/v2/acts/church_of_loujess~youtube-transcript-rag-scraper/run-sync-get-dataset-items?token=<YOUR_APIFY_API_TOKEN>" \-H "Content-Type: application/json" \-d '{"videoUrl": "https://www.youtube.com/watch?v=IELMSD2kdmk", "targetLanguage": "en", "chunkSizeWords": 200, "includeMetadata": true}'
Claude Desktop, Cursor, Windsurf & OpenAI MCP Configuration
{"mcpServers": {"apify-youtube-transcript-rag-scraper": {"url": "https://mcp.apify.com/?actors=church_of_loujess/youtube-transcript-rag-scraper","headers": {"Authorization": "Bearer <YOUR_APIFY_API_TOKEN>"}}}}
❓ Frequently Asked Questions
Am I charged if a YouTube video has no captions or subtitles?
No. When a video has no manual subtitles and no automatic ASR captions (for example, silent B-roll or music uploads), the Actor outputs a diagnostic record with "status": "no_transcript_available" and "billable": false, and skips the per-item dataset charge.
Can I pass both videoUrl and videoUrls in the same API call?
Yes. videoUrl is preserved for 1-to-1 drop-in compatibility with single-video workflows, while videoUrls accepts bulk lists. If both are provided, they are merged and deduplicated by videoId automatically.
What is the difference between segments and transcript in the output?
Both fields contain the exact same deduplicated {text, start, duration} array. We include both field names so existing LangChain, n8n, and Make templates expecting either segments or transcript work without remapping.
How does the 3-tier self-healing extraction engine work?
Tier 1 queries youtube-transcript-api caption tracks; if YouTube serves a consent gate or empty track, Tier 2 switches to Android InnerTube json3 caption endpoints; and Tier 3 falls back to yt-dlp subtitle track extraction plus YouTube oEmbed metadata.
Can it translate non-English videos into English (or vice versa)?
Yes. Set targetLanguage to any ISO 639-1 code ("en", "es", "de", "fr", "pt", "ja", etc.). If a native track in that language exists, it is preferred; otherwise YouTube's translated caption track is requested automatically.
What chunkSizeWords setting is best for RAG vector databases?
The default 200 words (~260–280 tokens) works great for hybrid dense+sparse retrieval in Pinecone, Qdrant, and pgvector. Increase to 300–400 words for long-context models, or lower to 100 words for fine-grained video clip search.
Is it legal to extract public data with YouTube Transcript Scraper?
YouTube Transcript Scraper only accesses publicly available web pages, feeds, and registries that require no private account login or authentication bypass. Keep in mind that extracted contact fields (such as public business emails, names, or usernames) can still qualify as personal data under frameworks like the GDPR or CCPA depending on jurisdiction. Always ensure you have a lawful basis (such as legitimate B2B interest) and consult Apify's guide on the legality of web scraping.
Can I integrate YouTube Transcript Scraper with Clay, n8n, Make, Zapier, or Google Sheets?
Yes. Every run outputs clean, flat-friendly fields alongside nested detail objects and integrates natively with Clay, n8n, Make, Zapier, Slack, HubSpot, Airtable, Google Sheets, Airbyte, LangChain, and CrewAI via Apify Integrations. You can also attach webhooks to trigger downstream workflows automatically the instant a run succeeds.
Can I call YouTube Transcript Scraper programmatically via the Apify API?
Yes. Use the official apify-client Python package or Node.js NPM package, or call the synchronous /run-sync-get-dataset-items REST endpoint directly. Full specifications are in the Apify API v2 reference.
YouTube Transcript Scraper not working or need a custom feature?
We actively maintain and monitor all 26 Actors and MCP Servers in our portfolio. If you encounter an edge case, unexpected output, or want to request a new field, please open a ticket on the Actor's Issues tab with your Run ID so we can resolve it rapidly.