YouTube Transcript Scraper – Bulk, Playlists, Channels & RAG avatar

YouTube Transcript Scraper – Bulk, Playlists, Channels & RAG

Pricing

$2.50 / 1,000 transcript delivereds

Go to Apify Store
YouTube Transcript Scraper – Bulk, Playlists, Channels & RAG

YouTube Transcript Scraper – Bulk, Playlists, Channels & RAG

RAG-ready YouTube captions and transcript scraper for bulk videos, channels and playlists. Timestamped text, SRT/VTT, citation chunks. Works from cloud servers. $0.0025 per delivered transcript; failures free.

Pricing

$2.50 / 1,000 transcript delivereds

Rating

0.0

(0)

Developer

Jacob Calmelat

Jacob Calmelat

Maintained by Community

Actor stats

0

Bookmarked

1

Total users

0

Monthly active users

8 days ago

Last modified

Features

Limited permissions

Share

YouTube Captions & Transcript Scraper for RAG – Bulk Videos, Playlists, Channels (SRT, VTT)

RAG-ready YouTube captions and transcripts in bulk: paste video links, a playlist, a channel or a search, and get one clean row per video with the plain text, timestamped segments, citation chunks with deep links, SRT and VTT subtitles, chapters and an approximate LLM token count.

You pay only for delivered transcripts: $0.0025 per transcript, no start fee. Videos without captions, private or deleted videos, invalid links and blocked requests are returned as free status rows.

Made by ShepHouse, an AI-operated venture. Built and tested with AI assistance; see "Tested" below for what was actually measured.

Try it free

Failures (no captions, private, blocked) are never charged. On the free Apify plan you can run a short playlist or a few video URLs and download CSV/JSON. Example input uses a public TED talk so you can see SRT, segments and RAG chunks immediately.

Sample OK row (shape):

{"videoId":"iG9CE55wbtY","status":"OK","title":"Do schools kill creativity? | Sir Ken Robinson | TED","language":"en","wordCount":3170,"approxTokens":4394,"text":"Good morning. How are you? …","charged":true}

What a full run gives you

One row per video. Example (real output from the private test run, long fields shortened):

{
"videoId": "iG9CE55wbtY",
"status": "OK",
"title": "Do schools kill creativity? | Sir Ken Robinson | TED",
"channel": "TED",
"lengthSeconds": 1203,
"language": "en",
"isAutoGenerated": false,
"languageFallbackUsed": false,
"wordCount": 3170,
"approxTokens": 4394,
"segmentCount": 427,
"text": "Good morning. How are you? (Audience) Good. It's been great, ...",
"segments": [
{
"start": 27.103,
"dur": 2.575,
"text": "Good morning. How are you?"
}
],
"srt": "1\n00:00:27,103 --> 00:00:29,678\nGood morning. How are you?\n...",
"chapters": [],
"charged": true
}

Failed rows look like {"videoId": "...", "status": "FAILED", "errorCode": "NO_CAPTIONS", "charged": false}.

Quick start (API / AI agents)

curl -X POST "https://api.apify.com/v2/acts/shephouse~youtube-transcript-scraper-bulk/run-sync-get-dataset-items?maxTotalChargeUsd=0.50" \
-H "Authorization: Bearer $APIFY_TOKEN" -H "Content-Type: application/json" \
-d '{"startUrls": ["https://www.youtube.com/@TED"], "maxVideosPerSource": 20, "languages": ["en"], "outputFormats": ["text", "srt"]}'

maxTotalChargeUsd caps what a run can cost you (here at most 100 transcripts × $0.0025).

Python: youtube-transcript-api-style client

If your code uses youtube-transcript-api and gets RequestBlocked/IpBlocked on a server, our open-source (MIT) client shephouse-yt keeps a familiar fetch() interface and runs this Actor under your Apify account (same price: $0.0025 per delivered transcript, failures free):

$pip install shephouse-yt
from shephouse_yt import YouTubeTranscriptApi
t = YouTubeTranscriptApi().fetch("PXbVn-f_wuM", languages=["en"]) # needs APIFY_TOKEN
print(t[0].text, t[0].start)

Using an MCP client (Claude Desktop, Cursor)? shephouse-yt-mcp exposes this Actor as an MCP tool (uvx shephouse-yt-mcp).

Input

  • Videos, playlists, channels or searches (one per line): https://www.youtube.com/watch?v=…, a bare 11-character ID, youtu.be/…, Shorts, https://www.youtube.com/playlist?list=…, @handle or a channel URL (newest uploads first), a search URL, or plain text (treated as a search).
  • Search queries: optional extra searches.
  • Preferred languages: e.g. en, de. In each language, manual captions win over auto-generated. If none match, the first available track is returned and languageFallbackUsed is true. availableLanguages lists every track.
  • Output formats: plain text is always included; add segments, SRT and/or VTT, or ragChunks, LLM/RAG-ready chunks of about 450 tokens. Each chunk has start/end seconds, approxTokens and a citeUrl (https://youtu.be/ID?t=SECONDS), so answers can cite the exact moment. No extra charge.
  • Split by chapters: if the description has a chapter list starting at 0:00, each chapter gets its own text.
  • Max videos per source / Max delivered transcripts: caps for cost control. Also use Max cost per run in Run options.

Output fields

videoId, url, status, title, channel, channelId, lengthSeconds, language, isAutoGenerated, languageFallbackUsed, availableLanguages, text, charCount, wordCount, approxTokens, segmentCount, segments, ragChunks, srt, vtt, chapters, source, charged. On failures: errorCode, errorMessage.

Error codes (never charged): NO_CAPTIONS, NO_MATCHING_CAPTIONS, VIDEO_UNAVAILABLE, VIDEO_UNPLAYABLE, LOGIN_REQUIRED, EMPTY_CAPTIONS, BLOCKED_BY_YOUTUBE, FETCH_FAILED, INVALID_INPUT, INVALID_OR_EMPTY_SOURCE.

Pricing

  • $0.0025 per delivered transcript (pay per event). No Actor start fee. Failed or skipped videos are free.
  • Example: 1,000 delivered transcripts = $5.00.
  • Free Apify plan: each run delivers up to 25 transcripts so you can check the output on your own videos; further videos are skipped and not charged. Paid Apify plans have no such cap, at the same price.

Use with AI agents, n8n, Make and the API

Call it via the Apify API, the official Apify integrations (n8n, Make, Zapier) or Apify's MCP server. approxTokens and chapters help with chunking for RAG. approxTokens is an estimate (about 4 characters per token for Latin script), not an exact tokenizer count.

Tested

Private test runs on Apify (2026-10-24 PT), before publication:

  • Final QA: 2 runs, 48 inputs across 6 channels, a playlist, URL and plain-text searches (English, Spanish, German, Japanese queries), direct links and negative cases. 41 of 41 caption-bearing videos delivered; charged events matched delivered rows exactly (25 + 16); every failure was free.
  • Correct free rows for: videos without captions (4), an unavailable ID, an invalid link, a non-existent playlist and a non-existent channel.
  • Long videos up to 5.3 hours: the last caption timestamp ends within 30 s of the video length. Manual and auto captions, language preference order and auto fallback were checked; SRT/VTT cue counts match the segments.
  • One temporary read timeout was retried automatically and recovered. This is a small sample, not a reliability guarantee: YouTube changes can temporarily break transcript extraction.

Limits and notes

  • Only existing captions (manual or auto-generated) are returned; there is no speech-to-text for videos without captions.
  • Age-restricted, private and members-only videos are not accessible.
  • Uses Apify residential proxy internally; the cost is included in the per-transcript price.
  • Use transcripts in line with YouTube's Terms and applicable copyright law.

Channel → vector database (3 steps)

  1. Fetch transcripts for a channel. Paste a channel URL or @handle into this Actor (or call the API). Enable ragChunks in Output formats. Each OK row includes plain text, timestamped segments, and RAG chunks with citeUrl deep links (https://youtu.be/ID?t=SECONDS). Failures (no captions, private, blocked) are free status rows.
  2. Chunk and cache locally. Download the dataset as JSON/CSV, or pipe it through our open-source client shephouse-yt. Keep the start/end/citeUrl fields so answers can cite the exact second.
  3. Load into a vector DB or search index. Export JSONL for Chroma, pgvector, Pinecone or Qdrant, or use BM25 (SQLite FTS5) with no embeddings. For a tested end-to-end kit (cache, chunking, search, optional embeddings, optional channel monitor + pgvector/Qdrant loaders):

Polar is merchant of record. The kits are ShepHouse products (AI-operated). The Actor itself stays pay-per-delivered-transcript on Apify; the kits are optional downloads that sit on top.

FAQ

Why does youtube-transcript-api fail on AWS, GCP, Azure or CI? YouTube often blocks datacenter IPs with RequestBlocked / IpBlocked. This Actor fetches captions through Apify residential proxy, so the same videos that fail from a cloud VM usually succeed here. Price stays $0.0025 per delivered transcript; blocked or caption-less videos are free.

Which languages and auto-captions are supported? Set Preferred languages (for example en, de). In each language, manual captions win over auto-generated. If none match, the first available track is returned and languageFallbackUsed is true. availableLanguages lists every track on the video. There is no speech-to-text for videos that have no captions at all.

What counts as a free (non-charged) video? Only rows with status: OK and a delivered transcript are charged. Videos without captions, private/deleted/unavailable videos, invalid links, empty caption tracks, blocked requests and skipped-by-cap videos return a status/error row with charged: false and cost $0.

Changelog

  • 0.1: first release (validated in private QA, see Tested).

More Actors from ShepHouse

ShepHouse is an AI-operated venture; all three Actors are built, tested and maintained with AI assistance.