YouTube Search to Transcripts — keyword in, full text out

Search YouTube and get the transcripts back in the same run. One call instead of two actors and glue code.

What it does

• Get transcripts for the top videos on any topic
• Build a RAG dataset from a YouTube search
• Research what creators say about a keyword
• Bulk YouTube transcripts by search term
• Turn a topic into a text corpus

Pass several queries at once. Every row is tagged with the query that found it.

Why this exists

Every other tool splits this in half. Search actors give you a list of video ids and stop. Transcript actors need ids you already have. So the normal workflow is: run a search actor, extract the ids, feed them into a transcript actor, join the results.

This does it in one run. "Transcripts of the top 200 videos about X" is a single call.

You only pay for transcripts

reason	Meaning	Charged
— (ok: true)	Transcript returned	Yes
no-captions	Video has no captions	No
unavailable	Deleted, private, or region-locked	No
no-results	Query matched nothing	No
search-blocked	Search request failed	No

Search results always contain some videos without captions. Those cost nothing.

Input

Field	Required	Description
queries	yes	Search terms as you would type them into YouTube
maxResultsPerQuery	no	Most relevant first. Default 50. Your cost ceiling.
language	no	Two-letter code. Default en.
includeSegments	no	Add timestamps. Default false.
maxConcurrency	no	1–25, default 10

Example input

{
  "queries": ["rag pipeline tutorial", "vector database comparison"],
  "maxResultsPerQuery": 100,
  "language": "en"
}

Output

Every row has the same fields whether it succeeded or failed, so you can select columns without branching. Failed rows are never charged.

Field	Type	Description
sourceQuery	string	The search query this row came from.
success	boolean	True when a transcript was returned. Failed rows are never charged.
videoId	string	YouTube's 11-character video identifier.
videoUrl	string	Canonical watch URL for the video.
title	string	The video's title as published.
channel	string	Display name of the channel that published the video.
channelId	string	YouTube channel identifier, stable across renames.
durationSeconds	number	Length of the video in seconds.
viewCount	integer	Views at the time of the run.
language	string	BCP-47 code of the caption track returned.
isAutoGenerated	boolean	True when the captions were machine-generated rather than human-authored.
segmentCount	integer	Number of timed segments in the transcript.
fullText	string	The complete transcript as one plain-text string. Null on failure.
segments	array	Timestamped transcript segments. Only present when includeSegments is enabled.
errorCode	string	Machine-readable failure reason. Null on success.
errorMessage	string	Human-readable explanation of the failure. Null on success.

Example — success

{
  "sourceQuery": "retrieval augmented generation",
  "success": true,
  "videoId": "jNQXAC9IVRw",
  "videoUrl": "https://www.youtube.com/watch?v=jNQXAC9IVRw",
  "title": "Me at the zoo",
  "channel": "jawed",
  "channelId": "UC4QobU6STFB0P71PMvOGN5A",
  "viewCount": 358194821,
  "durationSeconds": 19,
  "language": "en",
  "isAutoGenerated": false,
  "segmentCount": 3,
  "fullText": "All right, so here we are, in front of the elephants. The cool thing about these guys is that they have really, really, really long trunks.",
  "errorCode": null,
  "errorMessage": null,
  "segments": [
    {
      "start": 0.84,
      "duration": 3.2,
      "text": "All right, so here we are, in front of the elephants."
    },
    {
      "start": 4.04,
      "duration": 3.6,
      "text": "The cool thing about these guys is that they have"
    },
    {
      "start": 7.64,
      "duration": 2.9,
      "text": "really, really, really long trunks."
    }
  ]
}

Example — failure

A failure carries the same fields, so nothing downstream has to branch.

{
  "sourceQuery": "retrieval augmented generation",
  "success": false,
  "videoId": "s1CFmzZzO4c",
  "videoUrl": "https://www.youtube.com/watch?v=s1CFmzZzO4c",
  "title": "Board meeting recording",
  "channel": "Example Corp",
  "channelId": "UCzzz999",
  "viewCount": 412,
  "durationSeconds": 3120,
  "language": null,
  "isAutoGenerated": null,
  "segmentCount": null,
  "fullText": null,
  "errorCode": "no-captions",
  "errorMessage": "This video has no captions published, so there is no transcript to return."
}

Error codes

• no-captions
• unavailable
• blocked
• empty-transcript
• unparseable-input
• error
• invalid-query
• search-blocked
• no-results

Use it for

• Transcripts from a YouTube search — keyword in, full text out
• Topic research over video — build a corpus without picking videos first
• Interview and earnings call research — find the videos by keyword first
• RAG and embedding pipelines seeded from a query
• Literature reviews over video, competitive content analysis
• Training datasets on a subject

Related actors

If you need	Use
Specific videos you already have URLs for	YouTube Transcript Scraper
Every video on a channel	YouTube Channel Transcript Scraper
Every video in a playlist	YouTube Playlist Transcript Scraper
A channel's Shorts	YouTube Shorts Transcript Scraper
Comments instead of the spoken text	YouTube Comments Scraper

Use it from an AI agent (MCP)

This Actor is callable as a tool through the Apify MCP server, so Claude, ChatGPT, Cursor and VS Code can run it directly.

Add the server to your MCP client:

{
  "mcpServers": {
    "apify": {
      "url": "https://mcp.apify.com",
      "headers": {
        "Authorization": "Bearer <YOUR_APIFY_TOKEN>"
      }
    }
  }
}

Then ask for what you want in plain language — for example “search YouTube for this topic and transcribe the results” — and the agent calls cleanfeed/youtube-search-transcript-downloader with the right input. Every output field is described in the dataset schema, so the agent knows what it is getting back before it runs anything.

Call it from code

Python

from apify_client import ApifyClient

client = ApifyClient("<YOUR_APIFY_TOKEN>")

run = client.actor("cleanfeed/youtube-search-transcript-downloader").call(run_input={
    "queries": ["retrieval augmented generation"],
})

for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    if item["success"]:
        print(item)

JavaScript

import { ApifyClient } from 'apify-client';

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

const run = await client.actor('cleanfeed/youtube-search-transcript-downloader').call({
    queries: ["retrieval augmented generation"],
});

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

cURL

curl -X POST "https://api.apify.com/v2/acts/cleanfeed~youtube-search-transcript-downloader/run-sync-get-dataset-items?token=<YOUR_APIFY_TOKEN>" \
  -H 'Content-Type: application/json' \
  -d '{"queries": ["retrieval augmented generation"]}'

Limitations

• A residential proxy is required. YouTube serves playabilityStatus: ERROR to datacenter IP ranges while the identical request succeeds from a home connection — measured 12/12 success on residential against 0/12 on a cloud host with no proxy, and 7/12 through a datacenter proxy. The input defaults to residential; changing it will break most runs. Full method and per-environment figures are published in the reliability benchmark.
• Not every video has captions. Videos with captions disabled return errorCode: no-captions and are never charged. This is a property of the video, not a failure of the run.
• Search results are not stable. YouTube personalises and reorders results, so the same query run twice can return a different set.
• Results are capped by maxResultsPerQuery; this is not an exhaustive crawl of everything matching the query.

FAQ

How is this different from the transcript Actor?

You do not need to know the videos in advance. Pass a query, and the search and the transcription happen in one run — no glue code between two steps.

Are the results the same every time?

Not exactly. YouTube personalises and reorders search results, so repeated runs can return a different set. Cap with maxResultsPerQuery.

Do I need a YouTube API key?

No key, no OAuth and no quota.

What if a matched video has no captions?

It returns errorCode: no-captions and is not charged; the other results still come back.

Can I search in another language?

Yes — set language to prefer that caption track, and phrase the query in the target language.

Notes

Only publicly available caption data is collected. No login, no personal data, no video downloads.