YouTube AI Transcript Extractor

What is YouTube AI Transcript Extractor?

YouTube AI Transcript Extractor returns one dataset row per YouTube video with the full transcript and an AI analysis of it: a summary in the length you choose, topic tags, timestamped chapters, and optionally action items, sentiment and quotable highlights, alongside the video title, channel, publish date, duration, view count and thumbnail. Content marketers, researchers, podcast and interview analysts, and AI agents use it to understand long videos without watching them or pasting thousands of words into a prompt.

Why people use this transcript extractor

• Structured insight, not just text. aiSummary, aiTopics and aiChapters arrive as clean fields next to transcriptText.
• Chapters with timestamps. Each entry in aiChapters has startSeconds, a title and a one-line summary, ready for show notes or video descriptions.
• Pick only what you need. includeSummary, includeTopics, includeChapters, includeActionItems, includeSentiment and includeQuotes switch each AI field on or off.
• Cross-language output. transcriptLanguage picks the caption track and aiOutputLanguage sets the language of the AI fields, so a Spanish talk can get an English summary.
• Token transparency. aiModelUsed, aiInputTokens and aiOutputTokens show what each analysis used.
• Bring your own key. anthropicApiKey runs the AI step on your own quota.

Best fit for this Actor

• Best fit for summarizing interviews, webinars, tutorials, podcasts and talks into notes, chapters and tags.
• Designed for individual videos: watch URLs, youtu.be links, Shorts URLs or bare 11-character video IDs.
• When you only need raw transcripts, or want to start from a channel or search, use YouTube Transcript and Subtitle Data Scraper instead.
• To build the video list first, start with YouTube Channel Videos Scraper, then pass the video URLs into videoUrls here.

Example workflow: from a webinar backlog to show notes

A B2B marketing team starts with the URLs of their last ten webinars in videoUrls, with summaryLength set to medium and includeQuotes and includeActionItems on. The dataset returns a summary, topic tags, chapters, takeaways and quotable lines per video. They paste aiChapters into each YouTube description, turn aiQuotes into social posts, and feed aiActionItems into a recap newsletter. They then schedule a weekly run on new uploads and export the rows to their content calendar.

Input

{
  "videoUrls": ["https://www.youtube.com/watch?v=dQw4w9WgXcQ"],
  "includeSummary": true,
  "summaryLength": "medium",
  "includeTopics": true,
  "includeChapters": true,
  "includeActionItems": false,
  "includeSentiment": false,
  "includeQuotes": false,
  "transcriptLanguage": "en",
  "aiOutputLanguage": "English"
}

Field	Type	What it controls
videoUrls	array of strings	YouTube watch URLs, youtu.be links, Shorts URLs or 11-character video IDs. Required.
includeSummary	boolean	Add aiSummary. Default true.
summaryLength	string	short, medium or long. Default medium.
includeTopics	boolean	Add aiTopics. Default true.
includeChapters	boolean	Add timestamped aiChapters. Default true.
includeActionItems	boolean	Add aiActionItems. Default false.
includeSentiment	boolean	Add aiSentiment. Default false.
includeQuotes	boolean	Add aiQuotes. Default false.
transcriptLanguage	string	Caption track language as an ISO 639-1 code. Default en.
aiOutputLanguage	string	Language for the AI fields. Default English.
anthropicApiKey	string	Optional, stored as a secret. Your own key for the AI step.
proxyConfiguration	object	Proxy settings. Default uses residential Apify Proxy.

Output

One row per video.

Field	What it contains
videoId, videoUrl, title, thumbnail	Video identity
channelName, channelUrl	Channel
publishedAt, durationSeconds, viewCount	Video metadata
transcriptText, wordCount	Full transcript and its length
languageUsed, isAutoGenerated	Caption track used and whether it was auto-generated
aiSummary	Narrative summary
aiTopics	Topic tags
aiChapters	Chapters with startSeconds, title and summary
aiActionItems, aiSentiment, aiQuotes	Optional AI fields, when switched on
aiModelUsed, aiInputTokens, aiOutputTokens	AI model and token usage
transcriptStatus	success, or the reason no transcript was available
scrapedAt	Collection time

Illustrative row for a short tutorial video (transcript trimmed):

{
  "videoId": "abc123XYZ00",
  "videoUrl": "https://www.youtube.com/watch?v=abc123XYZ00",
  "title": "How to Price a Freelance Project in 5 Minutes",
  "channelName": "Example Studio",
  "channelUrl": "https://www.youtube.com/channel/UCexample000000000000000",
  "publishedAt": "2025-03-14T09:00:00-07:00",
  "durationSeconds": 312,
  "viewCount": 48210,
  "thumbnail": "https://i.ytimg.com/vi_webp/abc123XYZ00/maxresdefault.webp",
  "transcriptText": "Today I'll show you the three numbers I use to price every project...",
  "wordCount": 912,
  "languageUsed": "en",
  "isAutoGenerated": true,
  "aiSummary": "The creator explains a three-step method for pricing freelance work: estimate hours, set a target day rate, then add a risk buffer.",
  "aiTopics": ["Freelancing", "Pricing", "Project Estimates"],
  "aiChapters": [
    { "startSeconds": 0, "title": "Why pricing goes wrong", "summary": "Common reasons freelancers undercharge." },
    { "startSeconds": 74, "title": "The three numbers", "summary": "Hours, day rate and risk buffer explained." }
  ],
  "aiActionItems": null,
  "aiSentiment": null,
  "aiQuotes": null,
  "aiModelUsed": "claude-haiku-4-5-20251001",
  "aiInputTokens": 1976,
  "aiOutputTokens": 785,
  "transcriptStatus": "success",
  "scrapedAt": "2026-10-01T13:32:15.715Z"
}

Pricing

This Actor uses Pay per event plus Apify platform usage. Each video with a transcript is one Transcript extracted event, and each video that gets AI fields is one AI enrichment event; with your own anthropicApiKey the AI enrichment event is skipped. Every run also has one Actor Start event plus the platform usage of the run. For example, ten videos with summaries are ten Transcript extracted events, ten AI enrichment events and one start event. Videos without an available transcript add no transcript or AI events. The live Pricing tab on this page shows the current event prices.

API example

curl -X POST "https://api.apify.com/v2/acts/khadinakbar~youtube-ai-transcript-extractor/run-sync-get-dataset-items?token=YOUR_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"videoUrls": ["https://www.youtube.com/watch?v=dQw4w9WgXcQ"], "summaryLength": "short", "includeChapters": true}'

from apify_client import ApifyClient

client = ApifyClient("YOUR_API_TOKEN")
run = client.actor("khadinakbar/youtube-ai-transcript-extractor").call(run_input={
    "videoUrls": ["https://www.youtube.com/watch?v=dQw4w9WgXcQ"],
    "includeQuotes": True,
})
for video in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(video["title"], video["aiSummary"])
    for chapter in video.get("aiChapters") or []:
        print(chapter["startSeconds"], chapter["title"])

Use with AI agents (MCP)

Connect the Actor to Claude, ChatGPT or Cursor through Apify MCP and ask:

Summarize this YouTube interview in five sentences, list its chapters with timestamps, and pull three quotable lines.

The agent passes the URL in videoUrls with includeQuotes on, reads the row with get-dataset-items, and answers from aiSummary, aiChapters and aiQuotes, linking back to videoUrl. It checks transcriptStatus first, and each video costs one transcript event plus one AI event, so the cost of a request scales with the number of URLs.

Best results

Starting situation	Helpful action	Expected outcome
Show notes for a long video	Keep includeChapters on and set summaryLength to long	Detailed chapters and summary
Tutorials and how-tos	Turn on includeActionItems	Step-by-step takeaways
Social clips	Turn on includeQuotes	Quotable lines with context
A non-English video	Set transcriptLanguage to its language and aiOutputLanguage to yours	Translated analysis
High volume	Add your own anthropicApiKey	AI step billed to your own quota
Raw text only	Switch off every AI field	Transcript rows for your own pipeline

Builder's note

I built this because pasting a full transcript into a chat is slow and expensive, while a summary, tags and chapters answer most questions about a video. I made each AI field a separate switch so a run only produces, and pays for, the analysis you actually use, and I kept the transcript in every row so the AI output can always be checked against the source.

FAQ

What happens when a video has no captions?

The row's transcriptStatus names the reason, and only the start event applies to that video.

Does it work with Shorts?

Yes. Shorts URLs are handled like standard watch URLs.

Can I analyze a whole channel?

Collect the channel's video URLs first with the YouTube Channel Videos Scraper linked above, then pass them in videoUrls.

Responsible use

This Actor reads public YouTube captions and video metadata. Transcripts are the creators' content, so respect copyright and fair-use rules, attribute creators when you quote them, and follow YouTube's terms and applicable laws.