🎬 YouTube Video Summarizer — AI Summaries, Transcripts & Full Video Metadata

Turn any YouTube video into a short AI summary, a detailed markdown write-up, a timestamped transcript, and a complete metadata record — views, likes, comments, duration, tags, thumbnails, channel details and a calculated engagement rate.

Paste one URL or a hundred. No API key to configure, no transcript copy-pasting, no watching a 45-minute video to find out whether it is worth watching.

---

What you get

🧠 AI output	📜 Transcript	📺 Video & channel metadata
Concise summary	Timestamped segments	Title, description, video ID
Detailed markdown long-form summary	Start time and duration per segment	Views, likes, comment count
Summary source (transcript / title+description / basic)	Auto-caption extraction	Duration, resolution, category
	Language detection	Publication date + relative date
		Tags and thumbnails
		Channel name, handle, URL, ID
		Subscriber count, logo, verified badge
		Engagement rate (likes + comments ÷ views)
		Live status and upload type

---

Key features

• Two summaries per video. A short summary for scanning a list, and a long_summary in markdown for reading properly — so one run serves both triage and deep review.
• Honest summary provenance. summary_source tells you whether the summary came from the transcript, from the title and description (when no captions exist), or is a basic fallback. You always know how much to trust it.
• Real timestamped transcripts. Auto-captions are extracted into segments with start and dur values — ready for search, clipping or subtitle work.
• Complete metadata, not just the summary. Everything you would want from a YouTube scraper is in the same row, including a pre-calculated engagement_rate.
• Bulk processing. Add as many video URLs as you like; each is processed independently and pushed to the dataset as it completes.
• Proxy fallback built in. Starts on a direct connection and switches automatically to residential if YouTube blocks the request — including on "sign in to confirm you're not a bot" responses.
• Configurable retries. 1–10 attempts per URL, with exponential backoff for timeouts and connection errors.
• Nothing disappears. A video that cannot be processed still produces a row with status: "error" and the reason — unavailable, private, premium-only, or too long.

---

Use cases

• Content research — summarise dozens of videos in a niche and read only the ones that matter.
• Competitor analysis — see what competitors cover, how long their videos run and how well they engage.
• Learning & study — turn lectures and tutorials into readable notes with a timestamped transcript to jump back to.
• Newsletter and content curation — generate summaries for a weekly roundup automatically.
• SEO and keyword research — transcripts and tags reveal the language a niche actually uses.
• Podcast and interview digests — extract key points from long-form conversations.
• Market research — engagement rates across a topic show what genuinely resonates.
• Accessibility and repurposing — transcripts feed subtitles, blog posts and social clips.

---

How it works

1. You supply YouTube video URLs.
2. Each video's metadata is extracted — title, description, stats, tags, thumbnails and channel details.
3. Auto-captions are fetched where available and parsed into timestamped segments.
4. An AI summary is generated from the transcript; if there is no transcript, the title and description are used instead, and the fallback is recorded in summary_source.
5. On a block, the connection switches from direct to residential and the request is retried.
6. Each video is pushed to the dataset with status: "ok" or an explanatory error.

---

Quick start

1. Open the Actor and paste one or more YouTube URLs into YouTube Video URLs.
2. Leave the other settings at their defaults for a first run.
3. Click Start, then export the Output tab as JSON (recommended — transcripts are nested) or CSV.

Minimal input

{
  "urls": ["https://www.youtube.com/watch?v=dQw4w9WgXcQ"]
}

---

Input configuration

Field	Type	Default	Description
urls	array	— (required)	YouTube video URLs. Both https://www.youtube.com/watch?v=VIDEO_ID and https://youtu.be/VIDEO_ID work. Add many for bulk processing.
proxyConfiguration	object	direct	Optional. Leave empty to start with a direct connection; the Actor switches to residential automatically if YouTube blocks it. Enable Apify Proxy to use your preferred group first — recommended for large runs.
maxRetries	integer	3	Attempts per URL for transient failures, 1–10. Higher values are more resilient to temporary blocks and network issues.

Optional environment variables. Summarisation runs out of the box. If you prefer to use your own AI account, set OPENAI_API_KEY, or the Azure variables AZURE_OPENAI_ENDPOINT, AZURE_OPENAI_API_KEY, AZURE_OPENAI_DEPLOYMENT and AZURE_OPENAI_API_VERSION, in the Actor's environment settings.

---

Output data

One row per video.

AI summary

Field	Type	Description
summary	string	Concise overview of the video.
long_summary	string	Detailed summary in markdown.
summary_source	string	transcript, title_description, basic or none.

Video

id, video_id, url, title, description, published_date, relativeDate, duration_seconds, views, likes, comment_count, engagement_rate, tags, thumbnails, category, language, resolution, live_status, upload_type, type.

Transcript

transcript — an array of { start, dur, text } segments (seconds as strings), when captions are available.

Channel

channel.id, channel.name, channel.handle, channel.url, channel.subscriberCount, channel.logo, channel.badges (e.g. Verified), plus flat channel_id and channel_name.

Status

status — ok or error; error — the reason when processing failed; inputUrl — preserved on failed rows.

Example output (trimmed)

{
  "status": "ok",
  "video_id": "dQw4w9WgXcQ",
  "url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
  "title": "How We Scaled to 100k Users Without Paid Ads",
  "published_date": "2026-07-14",
  "relativeDate": "3 weeks ago",
  "duration_seconds": 1284,
  "views": 482100,
  "likes": 21400,
  "comment_count": 1820,
  "engagement_rate": 0.0481,
  "category": "Education",
  "language": "en",
  "resolution": "1920x1080",
  "tags": ["startup", "growth", "marketing"],
  "summary": "A walkthrough of three organic growth channels — SEO, community and partnerships — with the specific tactics and timelines behind reaching 100k users.",
  "long_summary": "# How We Scaled to 100k Users Without Paid Ads\n\n## Key points\n- SEO compounded after month 7…\n\n## Takeaways\n…",
  "summary_source": "transcript",
  "transcript": [
    { "start": "0.120", "dur": "3.480", "text": "When we started, we had no budget at all." },
    { "start": "3.600", "dur": "4.020", "text": "So everything had to be organic." }
  ],
  "channel": {
    "id": "UCexample123",
    "name": "Example Founders",
    "handle": "examplefounders",
    "url": "https://www.youtube.com/@examplefounders",
    "subscriberCount": "184000 subscribers",
    "badges": ["Verified"]
  }
}

Illustrative values — a live run returns real video data.

---

Usage examples

Bulk summarise a research list

{
  "urls": [
    "https://www.youtube.com/watch?v=VIDEO_ID_1",
    "https://youtu.be/VIDEO_ID_2",
    "https://www.youtube.com/watch?v=VIDEO_ID_3"
  ],
  "maxRetries": 3
}

Large run with proxy from the start

{
  "urls": ["…50 video URLs…"],
  "proxyConfiguration": { "useApifyProxy": true, "apifyProxyGroups": ["RESIDENTIAL"] },
  "maxRetries": 5
}

Build a transcript search index

Run across your target videos, then flatten the transcript arrays — each segment carries a start time, so you can link straight to the moment in the video.

Weekly competitor digest

Save a Task with your competitors' latest video URLs, attach a Schedule, and route summary into Slack or e-mail.

Filter by trustworthiness

Keep only rows where summary_source is transcript when you need summaries grounded in what was actually said.

---

Run it from your own code

Python

from apify_client import ApifyClient

client = ApifyClient("<YOUR_APIFY_API_TOKEN>")

run = client.actor("scraperx/youtube-video-summarizer").call(run_input={
    "urls": ["https://www.youtube.com/watch?v=dQw4w9WgXcQ"],
    "maxRetries": 3,
})

for v in client.dataset(run["defaultDatasetId"]).iterate_items():
    if v.get("status") == "ok":
        print(v["title"], "->", v["summary"])

JavaScript

import { ApifyClient } from 'apify-client';

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

const run = await client.actor('scraperx/youtube-video-summarizer').call({
    urls: ['https://www.youtube.com/watch?v=dQw4w9WgXcQ'],
});

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

cURL

curl -X POST "https://api.apify.com/v2/acts/scraperx~youtube-video-summarizer/runs?token=<YOUR_APIFY_API_TOKEN>" \
  -H "Content-Type: application/json" \
  -d '{"urls":["https://www.youtube.com/watch?v=dQw4w9WgXcQ"]}'

Integrations

Send summaries to Google Sheets, Airtable, Slack, Notion-compatible tools, Make, Zapier, Google Drive or your own endpoint via webhooks, and use Schedules for a recurring content digest.

---

Pricing

Pricing for this Actor is shown on its Pricing tab on the Actor page, and Apify displays the applicable cost before you run it. Free Apify accounts include monthly platform credit that is enough to trial it.

If a run escalates to Apify Residential proxies, that traffic is billed separately by the platform according to your plan.

---

Limits & good to know

• Maximum video length is 1 hour. Longer videos are rejected with a clear error rather than partially processed.
• Private, unavailable, premium-only and subscriber-only videos cannot be processed — they return status: "error" with the reason.
• No captions means a weaker summary. When a video has no auto-captions, the summary is built from the title and description, and summary_source says so.
• Transcripts come from YouTube's auto-captions, so accuracy depends on audio quality and language.
• Channel logo lookup is skipped for videos over 30 minutes to keep runs fast.
• Thumbnail URLs point at YouTube's CDN and can change.
• Use a proxy from the start on large runs; the automatic fallback handles occasional blocks, but pre-configuring is smoother at scale.
• Default run options are 4 GB memory and a 1-hour timeout — raise the timeout for long URL lists.

---

FAQ

Do I need an OpenAI API key? No. Summarisation works out of the box. If you would rather use your own account, set OPENAI_API_KEY or the Azure OpenAI variables in the Actor's environment settings.

Does it work on videos without subtitles? Yes, but the summary is then generated from the title and description. summary_source records this so you know.

Can I get the transcript on its own? Yes — the transcript array is included with every successful row, with start times and durations.

What is the maximum video length? One hour. Longer videos return an explanatory error.

Can I process many videos at once? Yes — add all the URLs to the input. Each is processed independently and streams into the dataset.

Why did a video fail? Check the error field: private, unavailable, premium-only, too long, or a block. Blocks are usually solved by enabling Apify Proxy and raising maxRetries.

What is engagement_rate? (likes + comments) ÷ views, calculated for you so you can rank videos immediately.

Which export format should I use? JSON — transcript, channel, tags and thumbnails are nested. CSV works if you only need summaries and metrics.

---

Legal & responsible use

This Actor processes only publicly available YouTube videos and their public auto-captions — the same content any viewer can access. It does not log in, download video files, or bypass paywalls or age restrictions. Video content, transcripts and thumbnails remain the property of their creators: AI summaries are derivative work, so credit the source, link to the original video, and comply with YouTube's Terms of Service and copyright law when republishing anything derived from them.

Support

Need longer video support, different summary formats, or a custom research pipeline? Open an issue on the Issues tab of this Actor.