YouTube Video Summarizer - Summary, Chapters & Blog Post

Get the substance of any YouTube video without watching it. Paste a link and receive a written summary, 5-8 key takeaways, topic tags, timestamped chapters, verbatim key quotes and, optionally, a complete blog post draft - all generated from the video's transcript. Handles hour-long podcasts, lectures and tutorials in a single pass.

Works on any public video with captions (manual or auto-generated) and outputs one clean row per video for JSON, CSV or Google Sheets.

What you get

For every video:

• summary - short, medium or long, written to be specific (names, numbers, tools, conclusions)
• keyTakeaways - 5-8 actionable points
• topics - 3-8 tags
• chapters - timestamp, start seconds, title and one-line summary for each section, ready to paste into a video description
• keyQuotes - 3-6 verbatim, timestamped quotes
• blogPost - optional 600-1000 word Markdown article based on the video
• transcript - optional plain-text transcript, plus transcriptLanguage, transcriptIsAutoGenerated, transcriptWordCount
• Video metadata: videoId, url, title, channelTitle, channelId, uploadDate, durationSeconds, viewCount
• model, promptTokens, completionTokens, status, error

Use cases

• Research and learning - digest lectures, conference talks and interviews in minutes
• Content repurposing - turn videos into blog posts, newsletters and social threads
• Chapter generation - add timestamps to your own uploads automatically
• Competitive monitoring - summarise every new video from rival channels on a schedule
• Knowledge bases - index a channel's back catalogue as searchable text
• Meeting and webinar notes - summarise recorded sessions published to YouTube

How to use

1. Enter one or more Video URLs (up to 50 per run).
2. Pick a Summary length and toggle Generate chapters, Extract key quotes, Write a blog post draft and Include full transcript.
3. Optionally set an Output language (for example Spanish) to translate the summary, and Transcript languages if the video is not in English.
4. Choose a Model - Gemini 2.5 Flash is the default and handles very long videos.
5. Run the actor - results appear in the Dataset tab.

Output format

{
  "inputUrl": "https://www.youtube.com/watch?v=8jPQjjsBbIc",
  "videoId": "8jPQjjsBbIc",
  "url": "https://www.youtube.com/watch?v=8jPQjjsBbIc",
  "title": "How to speak so that people want to listen | Julian Treasure | TED",
  "channelTitle": "TED",
  "channelId": "UCAuUUnT6oDeKwE6v1NGQxug",
  "uploadDate": "2014-06-27",
  "durationSeconds": 597,
  "viewCount": 50123456,
  "transcriptLanguage": "en",
  "transcriptIsAutoGenerated": false,
  "transcriptWordCount": 1560,
  "transcriptTruncated": false,
  "transcript": null,
  "summary": "Julian Treasure argues that most people are not listened to because of seven speaking habits ...",
  "keyTakeaways": ["Avoid the seven deadly sins of speaking: gossip, judging, negativity, complaining, excuses, exaggeration and dogmatism", "..."],
  "topics": ["public speaking", "communication", "voice"],
  "chapters": [
    { "timestamp": "0:00", "startSeconds": 0, "title": "Why nobody is listening", "summary": "The human voice is the most powerful instrument, yet most people are not heard." }
  ],
  "keyQuotes": [
    { "timestamp": "1:42", "startSeconds": 102, "quote": "The human voice: it's the instrument we all play." }
  ],
  "blogPost": null,
  "model": "google/gemini-2.5-flash",
  "promptTokens": 2870,
  "completionTokens": 910,
  "proxyUsed": "none",
  "status": "success",
  "error": null
}

Pricing

Each successfully summarised video is charged a small base fee plus a per-minute fee for the length of the video, so a 5-minute clip costs a fraction of a 3-hour podcast. Videos without a transcript, or that are private or removed, are reported with an error and are not charged.

Tips

• Very long videos (3+ hours) are summarised from the first ~160,000 characters of transcript; transcriptTruncated tells you when that happened.
• Chapters follow the creator's own chapter markers when present and correct them when the content disagrees.
• Choose Long summary plus Write a blog post draft for full repurposing; choose Short for scanning many videos cheaply.
• If transcripts fail with a bot-check message, set Proxy mode to Residential.

FAQ

Does it work without captions? No. The actor reads the video's manual or auto-generated captions. Videos with captions disabled return an error row.

Can it summarise in another language? Yes. Set Output language to any language name. The transcript can also be in a non-English language - list it under Transcript languages.

Is the summary accurate? The model is instructed to use only what is in the transcript and to be specific. Key quotes are verbatim so you can verify claims at the given timestamp.