🎬 YouTube Insight & Summarisation

Extract deep, structured insights from any YouTube video — without watching it. Summaries, timestamped key moments, and full transcripts. Works even when captions are missing. Supports Groq (free), Gemini (free), and OpenAI.

💰 Free to use — you only pay standard Apify compute costs. No per-result fees unlike most competitors.

---

Why YouTube Insight & Summarisation?

Most YouTube summarizers break when a video has no captions. This Actor doesn't.

It uses a two-step transcript pipeline:

1. Caption extraction first — fast, free, works on most videos
2. ASR audio transcription fallback — automatically kicks in when captions are unavailable, using AI speech recognition directly on the audio

This means it works on Hindi podcasts, regional content, older videos, and any video where the creator has disabled subtitles — content that other summarizers simply fail on.

---

Who Is This For?

• 🎓 Students & researchers — extract key insights from long lectures without watching in full
• 📰 Content creators — repurpose YouTube content into blogs, newsletters, or social posts
• 💼 Professionals — catch up on industry podcasts and interviews in minutes
• 🌍 Non-English speakers — get summaries in your native language using outputLanguage
• 🔁 Automation builders — integrate YouTube summarization into n8n, Zapier, or custom pipelines via API
• 🤖 Developers — call via REST API and plug structured JSON output into your product

---

Quick Start

Paste this into the Actor input and run:

{
  "urls": ["https://www.youtube.com/watch?v=UF8uR6Z6KLc"],
  "llmProvider": "groq",
  "openaiApiKey": "YOUR_GROQ_KEY_HERE",
  "includeTimestamps": true
}

💡 Get a free Groq key at console.groq.com — no credit card needed. Get a free Gemini key at aistudio.google.com — also free, higher token limits.

---

What This Actor Does

• Accepts one or more YouTube URLs (batch processing supported)
• Fetches video metadata: title, channel, description, views, duration, thumbnail
• Fetches transcript using the caption-first → ASR fallback pipeline
• Summarizes with your chosen LLM provider (Gemini, Groq, or OpenAI)
• Optionally returns:
  • ⏱ Timestamped key points with clickable deep-linked YouTube URLs
  • 📄 Full transcript text in timestamped format

---

Features

Feature	Details
✅ Works without captions	ASR fallback via agentx/video-transcript
⏱ Timestamped key points	MM:SS / HH:MM:SS with deep links (&t=Xs)
🌍 Multilingual output	Summarize in any language via outputLanguage
📦 Batch processing	Process multiple URLs in a single run
🔌 LLM flexibility	Groq (free), Gemini (free), or OpenAI
📄 Full transcript	Optional raw transcript output

---

Transcript Fallback Logic

For each video, the Actor follows this order:

Step 1 → Call pintostudio/youtube-transcript-scraper (caption-based, fast & free)
           ↓ captions found?
          YES → use captions, transcriptSource: "captions"
           NO ↓
Step 2 → Call agentx/video-transcript (ASR audio transcription, ~$0.43/video)
           ↓ ASR succeeded?
          YES → use ASR transcript, transcriptSource: "asr"
           NO ↓
Step 3 → Push error item, skip video, continue to next URL

The transcriptSource field in the output tells you exactly which method was used.

---

Input Schema

Field	Type	Required	Default	Description
urls	string[]	✅ Yes	—	List of YouTube video URLs to process
llmProvider	"gemini" | "groq" | "openai"	No	openai	LLM provider for summarization
openaiApiKey	string	✅ Yes	—	API key for the selected LLM provider
modelName	string	No	provider default	Override the default model
outputLanguage	string	No	same as video	Output language (e.g. Spanish, Hindi)
includeTimestamps	boolean	No	true	Generate timestamped key points
includeCommentAnalysis	boolean	No	true	Analyze top comments for sentiment
includeFullTranscript	boolean	No	false	Include full transcript in output
maxCommentsToAnalyze	number	No	50	Max comments to include in sentiment analysis

Example Input

{
  "urls": [
    "https://www.youtube.com/watch?v=UF8uR6Z6KLc",
    "https://www.youtube.com/watch?v=dQw4w9WgXcQ"
  ],
  "llmProvider": "groq",
  "openaiApiKey": "YOUR_GROQ_KEY_HERE",
  "outputLanguage": "",
  "includeTimestamps": true,
  "includeCommentAnalysis": true,
  "includeFullTranscript": false,
  "maxCommentsToAnalyze": 50
}

---

LLM Providers & Recommended Models

Provider	Default Model	Free Tier	Get Key
groq	llama-3.3-70b-versatile	✅ Yes (rate limited)	console.groq.com
gemini	gemini-2.0-flash	✅ Yes (generous limits)	aistudio.google.com
openai	gpt-4o-mini	❌ Paid	platform.openai.com

💡 Recommended for most users: Start with Groq for short videos (< 30 min). Use Gemini for longer content — it handles larger contexts on the free tier.

---

Output Schema

Successful result

{
  "videoUrl": "https://www.youtube.com/watch?v=UF8uR6Z6KLc",
  "videoId": "UF8uR6Z6KLc",
  "videoTitle": "Steve Jobs' 2005 Stanford Commencement Address",
  "channelName": "Stanford",
  "description": "...",
  "publishedAt": "2008-03-17",
  "viewCount": "45000000",
  "likeCount": "900000",
  "duration": "15m 4s",
  "thumbnailUrl": "https://img.youtube.com/vi/UF8uR6Z6KLc/maxresdefault.jpg",
  "summary": "Steve Jobs shares three personal stories...",
  "keyPoints": [
    {
      "timestamp": "1:00",
      "seconds": 60,
      "timestampedUrl": "https://www.youtube.com/watch?v=UF8uR6Z6KLc&t=60s",
      "topic": "Connecting the Dots",
      "summary": "Jobs dropped out of Reed College but audited classes that later shaped the Mac's typography."
    }
  ],
  "keyPointsCount": 8,
  "commentAnalysis": {
    "overallSentiment": "Positive",
    "consensusPoints": ["Highly motivational", "Timeless advice"],
    "controversialPoints": [],
    "topThemes": ["Career", "Purpose", "Death"],
    "summary": "Audience overwhelmingly finds the speech life-changing..."
  },
  "fullTranscript": null,
  "outputLanguage": "same as video",
  "modelUsed": "groq/llama-3.3-70b-versatile",
  "transcriptSource": "captions",
  "processedAt": "2026-04-24T12:00:00.000Z"
}

Failed result (both caption and ASR failed)

{
  "videoUrl": "https://www.youtube.com/watch?v=VIDEO_ID",
  "videoId": "VIDEO_ID",
  "error": "Transcription failed: video is private or region-locked",
  "processedAt": "2026-04-24T12:00:00.000Z"
}

---

Cost & Billing

• ✅ This Actor is free — no per-result charges
• You only pay standard Apify compute costs (typically a few cents per run)
• LLM token usage is billed directly by your provider (Groq and Gemini have free tiers)
• The ASR fallback (agentx/video-transcript) costs approximately $0.43 per video and is only triggered when captions are unavailable — videos with captions incur no ASR cost

---

Local Development

$npm install

Create or edit storage/key_value_stores/default/INPUT.json:

{
  "urls": ["https://www.youtube.com/watch?v=UF8uR6Z6KLc"],
  "llmProvider": "groq",
  "openaiApiKey": "YOUR_KEY_HERE",
  "includeTimestamps": true,
  "includeCommentAnalysis": true,
  "includeFullTranscript": false,
  "maxCommentsToAnalyze": 50
}

Run locally (Apify token required for sub-Actor calls):

$APIFY_TOKEN=your_apify_token npm run dev

Build check:

$npm run build

---

Deploy to Apify Console

npm install -g apify-cli
apify login
apify push

After deploying, run a smoke test with two videos:

• One with captions → expect transcriptSource: "captions"
• One without captions → expect transcriptSource: "asr"

---

Security

• Never commit your APIFY_TOKEN or LLM API keys to version control
• Use placeholder values in example inputs (e.g. "YOUR_GROQ_KEY_HERE")
• Rotate any keys that were accidentally shared or logged

---

License

ISC