YouTube Scraper — Videos, Comments & Transcripts

Extract videos, comments, and full transcripts from YouTube — no API key required. Works with channels, playlists, search queries, and direct video URLs. Built for AI/LLM datasets, market research, competitor analysis, and content intelligence.

---

💰 Pricing

Pay only for what you extract — three separate billing events:

What	Cost
📹 Videos	$15 per 1,000
💬 Comments	$8 per 1,000
📝 Transcripts	$5 per transcript

A small one-time actor-start fee applies per run. You are never charged for videos that are filtered out or skipped.

---

✨ Key Features

📝 Full Transcripts with Timestamps

Extracts complete captions — both manual and auto-generated — broken into timestamped segments. Each segment has a start time (seconds), duration, and text. The fullText field provides the entire transcript as a single string. Perfect for building AI training datasets, video summarization pipelines, search indexing, and subtitle analysis.

🔄 Four Input Types

• Channel URLs including @handle format — scrapes the full video library from a channel
• Playlist URLs — both user-created playlists (PL...) and YouTube Mix/Radio playlists (RD...)
• Search queries — finds videos matching one or more keyword queries
• Direct video URLs — pinpoint specific videos by URL or video ID

📊 Rich Video Metadata

Every video record includes: title, description, channel info, view count, like count, tags, category, duration in both human-readable and seconds format, publish date, thumbnail URL, and whether the video is a YouTube Short.

💬 Comments with Replies

Captures top-level comments and their replies in a single flat dataset. Each comment record includes the author name, channel ID, like count, reply count, pin status, and whether it is a reply (with parentCommentId for threading).

🔍 Powerful Filters

Filter videos before scraping details to save cost and time:

• Minimum views — skip low-traffic videos
• Date range — only videos published after a specific date
• Video type — regular videos only, Shorts only, or all

---

🚀 Quick Start

Option 1 — Channel URL

Paste a channel URL to scrape all videos from that channel.

https://www.youtube.com/@MrBeast
https://www.youtube.com/channel/UCX6OQ3DkcsbYNE6H8uQQuVA

Option 2 — Playlist

Paste a playlist URL. Works with standard playlists and YouTube Radio/Mix playlists.

https://www.youtube.com/playlist?list=PLxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
https://www.youtube.com/watch?v=dQw4w9WgXcQ&list=RDdQw4w9WgXcQ

Option 3 — Search Queries

Provide one or more search terms as a list. The scraper returns the top results for each query.

["python tutorial", "machine learning 2024", "startup advice"]

Option 4 — Direct Video URL

Provide one or more video URLs to scrape specific videos.

https://www.youtube.com/watch?v=dQw4w9WgXcQ
https://youtu.be/dQw4w9WgXcQ

---

⚙️ Input Parameters

Parameter	Type	Default	Description
startUrls	array	—	YouTube URLs: video, channel, playlist, or search results page
searchQueries	array	—	Keywords to search for (each query runs independently)
maxVideos	integer	50	Maximum number of videos to process per run
scrapeComments	boolean	false	Extract comments for each video
maxCommentsPerVideo	integer	100	Maximum comments to extract per video
scrapeTranscripts	boolean	false	Extract full transcripts (captions) with timestamps
filterVideoType	string	all	Filter by type: all, videos (excludes Shorts), or shorts
filterByMinViews	integer	0	Skip videos with fewer than this many views
filterByDateFrom	string	—	Only include videos published on or after this date (YYYY-MM-DD)
filterByLanguage	string	en	Preferred transcript language code (e.g. en, es, de)

---

📦 Output Format

All results are saved to the Apify dataset. Three record types are mixed in a single dataset and can be filtered by the type field.

---

Video Record (type: "video")

One record per video with full metadata.

{
  "type": "video",
  "videoId": "dQw4w9WgXcQ",
  "url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
  "title": "Rick Astley - Never Gonna Give You Up (Official Video) (4K Remaster)",
  "description": "The official video for \"Never Gonna Give You Up\" by Rick Astley...",
  "channelId": "UCuAXFkgsw1L7xaCfnd5JJOw",
  "channelName": "Rick Astley",
  "channelUrl": "https://www.youtube.com/channel/UCuAXFkgsw1L7xaCfnd5JJOw",
  "publishedAt": "2009-10-25",
  "duration": "3:33",
  "durationSeconds": 213,
  "viewCount": 1760000000,
  "likeCount": 18000000,
  "commentCount": 0,
  "thumbnailUrl": "https://i.ytimg.com/vi/dQw4w9WgXcQ/maxresdefault.jpg",
  "tags": ["rick astley", "never gonna give you up", "80s music", "rickroll"],
  "category": "Music",
  "isShort": false,
  "scrapedAt": "2026-04-09T07:32:21.574Z"
}

Field reference:

Field	Type	Description
videoId	string	YouTube video ID
url	string	Full watch URL
title	string	Video title
description	string	Full video description
channelId	string	YouTube channel ID (UC...)
channelName	string	Channel display name
channelUrl	string	Channel URL
publishedAt	string	Publish date (ISO format when available, human-readable otherwise)
duration	string	Duration in H:MM:SS or M:SS format
durationSeconds	integer	Duration in seconds
viewCount	integer	Total view count
likeCount	integer	Total like count
commentCount	integer	Comment count (may be 0 if not returned by the API)
thumbnailUrl	string	Max-resolution thumbnail URL
tags	array	Video tags set by the uploader
category	string	YouTube category (Music, Education, etc.)
isShort	boolean	True if video duration is 60 seconds or less
scrapedAt	string	ISO timestamp of when the record was created

---

Comment Record (type: "comment")

One record per comment or reply. Top-level comments and replies are saved as flat records — use isReply and parentCommentId to reconstruct threads.

{
  "type": "comment",
  "commentId": "UgxHPK_QyTdcuyvuX7B4AaABAg",
  "videoId": "dQw4w9WgXcQ",
  "videoTitle": "Rick Astley - Never Gonna Give You Up (Official Video) (4K Remaster)",
  "text": "I came here voluntarily. Nobody rickrolled me.",
  "authorName": "@username",
  "authorChannelId": "UCxxxxxxxxxxxxxxxxxxxxxxxxx",
  "likeCount": 134000,
  "replyCount": 463,
  "isReply": false,
  "parentCommentId": null,
  "publishedAt": "6 years ago",
  "isPinned": false,
  "scrapedAt": "2026-04-09T07:32:24.880Z"
}

Field reference:

Field	Type	Description
commentId	string	Unique YouTube comment ID
videoId	string	Parent video ID
videoTitle	string	Parent video title
text	string	Full comment text
authorName	string	Author's @handle
authorChannelId	string	Author's channel ID
likeCount	integer	Number of likes on the comment
replyCount	integer	Number of replies (top-level comments only)
isReply	boolean	True if this is a reply to another comment
parentCommentId	string	ID of the parent comment if isReply is true
publishedAt	string	Relative publish time as shown on YouTube
isPinned	boolean	True if the comment is pinned by the channel

---

Transcript Record (type: "transcript")

One record per video. Contains the full transcript broken into timed segments plus a concatenated fullText field.

{
  "type": "transcript",
  "videoId": "dQw4w9WgXcQ",
  "videoTitle": "Rick Astley - Never Gonna Give You Up (Official Video) (4K Remaster)",
  "language": "en",
  "languageName": "English (auto-generated)",
  "isAutoGenerated": true,
  "segments": [
    { "start": 18.8,  "duration": 7.16, "text": "We're no strangers to love." },
    { "start": 25.96, "duration": 4.32, "text": "You know the rules and so do I." },
    { "start": 30.28, "duration": 9.28, "text": "You wouldn't get this from any other guy." }
  ],
  "fullText": "We're no strangers to love. You know the rules and so do I...",
  "wordCount": 291,
  "scrapedAt": "2026-04-09T07:32:26.316Z"
}

Field reference:

Field	Type	Description
language	string	ISO 639-1 language code (e.g. en)
languageName	string	Human-readable language name
isAutoGenerated	boolean	True if auto-generated captions, false if manually created
segments	array	Array of timed segments with start, duration, and text
fullText	string	All segment text concatenated into one string
wordCount	integer	Total word count of the transcript

---

🔍 Use Case Examples

Build an LLM training dataset from a podcast channel

{
  "startUrls": [{ "url": "https://www.youtube.com/@lexfridman" }],
  "maxVideos": 200,
  "scrapeTranscripts": true,
  "filterByLanguage": "en",
  "filterVideoType": "videos"
}

Competitor sentiment analysis via comments

{
  "searchQueries": ["notion app review", "notion vs obsidian 2024"],
  "maxVideos": 50,
  "scrapeComments": true,
  "maxCommentsPerVideo": 300,
  "filterByMinViews": 10000
}

Scrape only YouTube Shorts from a channel

{
  "startUrls": [{ "url": "https://www.youtube.com/@MrBeast" }],
  "maxVideos": 200,
  "filterVideoType": "shorts"
}

Monitor new uploads (weekly incremental run)

{
  "startUrls": [{ "url": "https://www.youtube.com/@ycombinator" }],
  "maxVideos": 20,
  "filterByDateFrom": "2026-04-01"
}

Research trending videos with minimum reach

{
  "searchQueries": ["AI tools 2026", "ChatGPT tutorial"],
  "maxVideos": 30,
  "filterByMinViews": 500000,
  "filterByDateFrom": "2026-01-01",
  "scrapeComments": true,
  "maxCommentsPerVideo": 100
}

Scrape a specific playlist with transcripts

{
  "startUrls": [{ "url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ&list=RDdQw4w9WgXcQ" }],
  "maxVideos": 20,
  "scrapeTranscripts": true
}

---

📊 Who Uses This

Use Case	Who	What They Get
LLM / AI training data	AI & ML teams	Timestamped transcripts from thousands of videos in structured JSON
Competitive research	Product teams	What real users say about competitor products in comments
SEO & content strategy	Marketers	Top-performing titles, tags, view/engagement metrics
Influencer analytics	Agencies	Channel video performance, engagement trends over time
Shorts trend analysis	Creators	Trending Shorts topics, view/engagement ratios
Academic research	Researchers	Comment corpora for NLP, sentiment analysis, discourse studies
Market research	Analysts	Consumer opinions from product reviews and unboxing videos
Podcast transcription	Journalists & writers	Full text of long-form interviews and podcast episodes
Education data	EdTech companies	Course content from educational channels in structured format

---

💡 Pro Tips

1. Transcripts for AI training Use scrapeTranscripts: true with a focused channel. Set filterVideoType: "videos" to skip Shorts, which rarely have quality transcripts. Prefer channels that publish manual (not auto-generated) subtitles for higher accuracy.

2. Trending Shorts discovery Combine filterVideoType: "shorts" + filterByDateFrom (last 7–14 days) + filterByMinViews: 50000 to surface only breakout Shorts from the past week.

3. Comment sentiment at scale Run with scrapeComments: true and maxCommentsPerVideo: 500 on product review videos. Filter by filterByMinViews: 5000 to ensure the comments represent real audience volume.

4. Incremental monitoring Schedule this actor weekly. Set filterByDateFrom to the previous Monday. This way you only process newly published videos and never duplicate data.

5. Multi-query keyword research Pass 5–10 related search queries in searchQueries. Analyze the tags and title fields across results to discover the vocabulary your target audience uses.

6. Combining sources You can mix startUrls and searchQueries in a single run. Results from all sources are deduplicated — each video is processed only once even if it appears in multiple sources.

---

⚠️ Limits & Notes

• No API key required — the scraper uses YouTube's internal API, the same one used by the YouTube website and mobile apps.
• Transcripts availability — not all videos have captions. If no transcript is found, the video record is still saved, but no transcript record is created.
• Tags and category — some videos, particularly older ones or those on music label channels, do not return tags or category via the API. These fields will be empty arrays or empty strings in that case.
• publishedAt format — for most videos this is an ISO date. For some videos the YouTube API returns a relative string like "3 years ago" or "Premiered Dec 27, 2022" — these are preserved as-is.
• Comments — the scraper fetches comments sorted by Top Comments. Replies are included when paginating through comment threads.
• Rate limiting — the scraper includes automatic retry logic and pacing to handle YouTube's rate limits gracefully.

---

❓ FAQ

Q: Do I need a YouTube API key? No. The scraper uses YouTube's internal Innertube API — the same one used by the YouTube website and mobile apps — so no API key or developer account is required.

Q: How many videos can I scrape per run? There is no hard limit beyond maxVideos. You can set it to 1,000 or more. Keep in mind that each video requires a few API calls, so large runs take time. For channels or playlists with thousands of videos, consider running multiple incremental jobs with filterByDateFrom.

Q: Why is commentCount always 0 in the video record? YouTube's Innertube API does not reliably return the comment count in the metadata response. The actual comments are fetched separately when scrapeComments: true. The commentCount field is kept in the schema for compatibility but will typically be 0.

Q: Why does some videos have an empty description or missing publishedAt? Certain videos — especially on music label or VEVO channels — are restricted by YouTube's internal API and return limited metadata. The scraper makes a best-effort attempt to fill all fields using multiple API endpoints, but some fields may remain empty for these videos.

Q: What happens if a video has no captions? If scrapeTranscripts: true and no captions are available, no transcript record is created for that video. The video record is still saved. The scrapeTranscripts billing event only fires when a transcript is successfully retrieved.

Q: Can I scrape private or age-restricted videos? No. The scraper only accesses publicly available content — the same content visible to any logged-out user. Private, unlisted, or age-restricted videos are not accessible.

Q: How does billing work exactly? You are charged three separate per-event fees:

• $15 per 1,000 videos processed (billed per video, after filters pass)
• $8 per 1,000 comments extracted
• $5 per transcript successfully retrieved

A small one-time actor-start fee also applies. Videos filtered out by filterByMinViews, filterByDateFrom, or filterVideoType are not billed.

Q: Can I run this on a schedule? Yes. Use the Apify scheduler to run the actor weekly or daily. Set filterByDateFrom to the start of the period you want to capture. This avoids re-processing already-scraped videos and minimizes cost.

Q: How do I get transcripts in a specific language? Set filterByLanguage to the ISO 639-1 code (e.g. "en", "es", "de"). If a transcript in that language is not available, the scraper will fall back to any available language. To strictly require a language, filter the output by the language field in the transcript record.

Q: Are YouTube Shorts supported? Yes. Use filterVideoType: "shorts" to scrape only Shorts, filterVideoType: "videos" to exclude them, or filterVideoType: "all" (default) to include both.

---

⚖️ Legal & Ethical Use

This scraper accesses publicly available data on YouTube — the same data visible to any user without logging in. Use it for legitimate research, content analysis, and data science purposes.

Always comply with:

• YouTube Terms of Service
• Applicable data protection regulations (GDPR, CCPA, etc.)

Do not use scraped data to build spam systems, harass individuals, or circumvent YouTube's monetization systems.

---

🛠️ Technical Notes

• Built on the Apify SDK with pay-per-event billing (Actor.charge())
• Uses YouTube's Innertube API (the same internal API used by youtube.com) — no official Data API quota required
• Transcript fetching uses the Android Innertube client to access caption tracks without authentication tokens
• Comment pagination follows the 2024–2026 YouTube comment format using commentEntityPayload from frameworkUpdates.entityBatchUpdate.mutations
• Residential proxy is used automatically for all requests to ensure reliable access