YouTube Shorts Scraper: Comments & Creator Engagement Data

A YouTube Shorts scraper that pulls full video metadata from any channel, handle, or keyword search and the top and newest comments (with replies) on every Short -- including real creator-heart and pinned-comment flags. Use it as a YouTube Shorts data export tool, a YouTube comments scraper, or a lightweight creator-engagement analytics feed, no login required.

What this actor does

• Scrapes YouTube Shorts from a channel handle, a full channel URL, a plain username, or a free-text keyword search -- one run can process many inputs at once.
• Extracts full video metadata: title, view count, likes, publish date, duration, hashtags, description links, caption/subtitle availability, video format list, and channel details (avatar, banner, subscriber count, verification status, total videos/views, join date, location).
• Collects comments & replies as their own dataset rows for every Short, continuation-paginated so you're not capped at whatever loads on the first page.
• Surfaces real creator-engagement signals per comment: whether the video's creator hearted that specific comment, and whether the comment is pinned -- both read from the actual signal YouTube uses internally, not a guess.
• Optional AI sentiment & topic tagging on comment text, using your own AI provider API key (off by default).
• Filters by publish date, with an absolute date or a relative window of any length (days/weeks/months/years), and sorts by newest, oldest, or most popular.

Key features

• Flexible input: channel handles (@MrBeast), plain usernames (MrBeast), full channel URLs, or keyword search (funny shorts).
• Accurate result counts: ask for 3 Shorts, get exactly 3 -- the count you enter is what you get, not padded to a hidden minimum.
• Real date filtering: YYYY-MM-DD absolute dates or <N> day/week/month/year(s) relative windows for any N, matching what the input field actually promises.
• Comments & replies child rows: type: "comment" / type: "reply" rows linked to their parent video by videoId, each with author, text, like count, reply count, and the creator-heart/pinned flags above.
• Reliable connection handling: the run automatically adjusts its own connection settings if a request runs into trouble, so a single hiccup doesn't fail the whole run.
• Real-time output: every row -- videos and comments alike -- lands in the dataset as soon as it's collected, so you can watch results fill in live.
• 43 video-level fields + 17 comment-level fields, all visible in dedicated dataset table views (a "Scraped Shorts Videos" view and a separate "Comments & Replies" view).

Input

Example input

{
  "startUrls": ["mrbeast", "@MrBeast", "https://www.youtube.com/@MrBeast/shorts", "funny shorts"],
  "maxResults": 10,
  "publishedAfter": "30 days",
  "sortOrder": "newest",
  "fetchComments": true,
  "maxCommentsPerVideo": 20,
  "fetchReplies": false,
  "enableAiSentiment": false,
  "proxyConfiguration": { "useApifyProxy": false }
}

Input fields

Field	Type	Default	Description
startUrls	array (required)	["rainbowicecream9780"]	Channel handles, usernames, full channel URLs, or search keywords.
maxResults	integer	10	Max Shorts per channel/keyword (1-10,000). Honored exactly.
publishedAfter	string	(none)	YYYY-MM-DD absolute date, or <N> day/week/month/year(s) relative (e.g. 17 days, 10 years).
sortOrder	enum	newest	newest, popular (by views), or oldest.
fetchComments	boolean	true	Collect top+newest comments for every scraped Short as separate dataset rows.
maxCommentsPerVideo	integer	20	Max combined comment + reply rows per Short (0-500).
fetchReplies	boolean	false	Expand replies under their parent comment (still capped by maxCommentsPerVideo).
enableAiSentiment	boolean	false	Tag each comment with sentiment + topic keywords using your own AI key.
aiApiKey	string (secret)	(none)	Your own API key for the AI model below. Only used when enableAiSentiment is on.
aiModel	string	(none)	Model name, e.g. gpt-4o-mini, claude-sonnet-4-5, gemini-2.5-flash.
proxyConfiguration	object	{ "useApifyProxy": false }	Standard Apify proxy configuration.

Output

Video row example

{
  "type": "shorts",
  "title": "Giving Trick-or-Treaters Weirder and Weirder Gifts",
  "id": "R7MkZL5i3FQ",
  "url": "https://www.youtube.com/shorts/R7MkZL5i3FQ",
  "thumbnailUrl": "https://i.ytimg.com/vi/R7MkZL5i3FQ/hqdefault.jpg",
  "viewCount": 64493572,
  "date": "2025-11-07",
  "likes": 909841,
  "channelName": "MrBeast",
  "channelUrl": "https://www.youtube.com/@MrBeast",
  "channelUsername": "MrBeast",
  "numberOfSubscribers": 450000000,
  "isChannelVerified": true,
  "duration": "0:27",
  "commentsCount": 7171,
  "commentsTurnedOff": false,
  "hashtags": ["shorts"],
  "isAgeRestricted": false,
  "isPaidContent": false,
  "isMembersOnly": false
}

Every video row carries the full field set: title, translatedTitle, type, id, url, thumbnailUrl, viewCount, date, likes, location, channelName, channelUrl, channelUsername, collaborators, channelId, channelDescription, channelJoinedDate, channelDescriptionLinks, channelLocation, channelAvatarUrl, channelBannerUrl, channelTotalVideos, channelTotalViews, numberOfSubscribers, isChannelVerified, inputChannelUrl, isAgeRestricted, aboutChannelInfo, duration, commentsCount, text, translatedText, descriptionLinks, subtitles, order, commentsTurnedOff, fromYTUrl, isMonetized, hashtags, formats, isMembersOnly, input, fromChannelListPage, isPaidContent.

Comment row example

{
  "type": "comment",
  "isChild": true,
  "videoId": "R7MkZL5i3FQ",
  "videoUrl": "https://www.youtube.com/shorts/R7MkZL5i3FQ",
  "commentId": "UgxABC...",
  "isReply": false,
  "text": "This made my day!",
  "author": "@somefan",
  "voteCount": 1240,
  "replyCount": 3,
  "hasCreatorHeart": true,
  "commentIsPinned": false,
  "authorIsChannelOwner": false,
  "authorIsVerified": false,
  "publishedTimeText": "2 days ago",
  "sentiment": null,
  "topics": null
}

Comment/reply rows carry: type, isChild, videoId, videoUrl, channelName, commentId, parentCommentId, isReply, text, author, authorAvatarUrl, authorIsChannelOwner, authorIsVerified, publishedTimeText, voteCount, replyCount, hasCreatorHeart, commentIsPinned, sentiment, topics, order, scrapedAt. sentiment/topics stay null unless enableAiSentiment is on and classification succeeds -- never a guessed value.

Use cases

• Creator engagement research: see which comments a creator actually hearted or pinned, not just raw comment counts.
• Content & trend research: analyze Shorts performance across channels or by keyword.
• Competitor monitoring: track competitor channels' Shorts output and audience reaction.
• Community sentiment tracking: pair the comments feed with the optional AI tagging to gauge audience reaction to a video or a channel over time.
• Dataset building: export a structured YouTube Shorts + comments dataset to CSV/JSON for downstream analysis.

Frequently asked questions

Q: Can I scrape private channels? A: No, the actor only scrapes publicly available Shorts and their public comments.

Q: What happens if a channel has no Shorts? A: The actor returns no rows for that input and continues with the rest.

Q: Is hasCreatorHeart a real signal or a guess? A: It's read from the actual per-comment engagement data YouTube returns, not inferred from anything else -- it's only true when the video's own creator hearted that specific comment.

Q: Are comments capped per video? A: Yes -- maxCommentsPerVideo caps the combined comment + reply rows per Short (default 20, max 500). Set it to 0 to skip comments even if fetchComments is on.

Q: What date filters are supported? A: An absolute YYYY-MM-DD date, or a relative window like 3 days, 17 days, 2 weeks, 1 month, or 10 years -- any whole number works, not just a couple of hardcoded values.

Q: Do I need my own AI key for sentiment tagging? A: Yes. enableAiSentiment is off by default; when you turn it on, supply your own API key in aiApiKey and a model name in aiModel. Comments are still collected normally either way.

Q: How are results sorted? A: By sortOrder: newest (default), oldest, or popular (by view count).

Cautions

• Data is collected only from publicly available YouTube pages.
• No data is taken from private accounts, password-protected content, or authenticated-only surfaces.
• The end user is responsible for ensuring legal compliance (privacy, data protection, platform terms) for their use case.