YouTube Summarizer

About YouTube Summarizer

YouTube Summarizer extracts and condenses the content of any YouTube video. It fetches the video's captions (auto-generated or manual), detects the language, and produces a concise summary using extractive summarization that runs locally in your browser.

The tool supports multiple output formats including bullet points for quick overviews, executive summaries for professionals, study notes for students, SEO outlines for content creators, social media captions, and full blog articles.

It also generates timestamped video chapters and extracts key terms and topics using TF-IDF analysis. You can export the full result to PDF, TXT, Markdown, or structured JSON for use in other apps and scripts.

How does YouTube Summarizer work?

The tool fetches the video's subtitle/caption track from YouTube (either auto-generated or manually added captions). It then uses the franc library to detect the language and applies a TF-IDF analysis to find the most important terms.

The summary itself is produced by extractive summarization: the transcript is split into sentences, each sentence is ranked by signals such as position, length, and keyword density, and the highest-ranked sentences are selected and re-ordered into a readable summary. All of this runs locally in your browser. Chapters are generated by splitting the timeline into segments and labeling each one with its top keywords.

What video types are supported?

YouTube Summarizer works with any YouTube video that has captions available:
- Regular YouTube videos
- YouTube Shorts
- Live streams (after they end)
- Embedded videos

The video must have either auto-generated captions or manually added subtitles. Videos without any captions cannot be summarized.

What output formats are available?

Six different output formats are available:

1. Bullet Points: Quick, scannable list of key points
2. Executive Summary: Professional summary with context
3. Study Notes: Formatted for learning with key terms highlighted
4. SEO Outline: Optimized for content creation with headers and keywords
5. Social Media Caption: Short, engaging summary with hashtags
6. Blog Article: Full article with sections and detailed breakdown

What languages are supported?

The tool automatically detects the video's language and can summarize most caption languages. Extractive summarization and keyword extraction work best for languages that separate words with spaces (such as English, Spanish, Portuguese, French, and Vietnamese). Languages without spaces between words, like Chinese, Japanese, or Thai, are still detected and summarized, but keyword and chapter-title quality may be lower.

The interface itself is available in English, Spanish, Vietnamese, Portuguese, and French.

How accurate are the chapters?

Chapters are generated heuristically: the transcript timeline is divided into roughly even segments (up to 8 per video), and each segment is labeled with its most frequent keywords. Timing is based on caption timestamps, not on detected topic changes, so chapter boundaries are approximate rather than semantic.

Well-structured videos with clear, evenly paced topics produce the most useful chapter breakdowns.

Can I export the summary?

Yes. You can export your result in four formats:
- PDF: professional document format, great for sharing
- TXT: plain text for easy editing
- Markdown: for blogs, GitHub, or note-taking apps
- JSON: structured, machine-readable output

The PDF, TXT, and Markdown exports include the summary, chapters, and keywords as readable prose.

Can I get machine-readable (JSON) output?

Yes. The Export JSON button downloads a single structured file containing the video metadata, detected language, word statistics and compression ratio, the ranked keywords with their scores and counts, the topics, the timestamped chapters (start, end, title, excerpt), the chosen format, and the full summary and transcript.

This makes the tool a data source you can feed directly into a CMS, spreadsheet, note app, or your own scripts and pipelines, with no copy-pasting from prose.

Is my data private?

Only the YouTube URL is sent to our server, and only so it can fetch the video's captions (YouTube does not allow the browser to retrieve them directly). The transcript text is then returned to your browser, where summarization, keyword extraction, language detection, and chapter generation all run locally. The summary and exports are built on your device and are not stored on our server.

How accurate is the summary?

The summary is extractive, not generative: it selects and re-orders the most representative sentences that already exist in the transcript rather than writing new text. This keeps it factual and free of hallucinations, but it works best on videos with clean, well-punctuated captions.

Auto-generated captions with no punctuation, heavy slang, or background noise will produce a rougher summary, and chapter timing is heuristic. Always review the output before relying on it for important work.