YouTube Comments Scraper

💬 YouTube Comments Scraper

Pull the comments from any public YouTube video into a clean dataset: text, author, likes, reply count, creator hearts, and publish time. Paste video links, pick how many comments you want, and export to JSON, CSV, or Excel. No API key, no login. $0.0012 per comment, which is $1.20 per 1,000.

✨ Why use YouTube Comments Scraper?

• 🔑 No API key or quota — you do not need a Google Cloud project or deal with daily quota limits.
• 🔗 Any video link — watch, youtu.be, shorts, embed, and live links all work, and you can mix them in one run.
• 🔥 Newest or top — choose newest first or top comments, or keep only comments after a date.
• 🧾 Rich rows — likes, reply count, channel-owner flag, creator heart, and the video title and total comment count on every row.
• 🚦 Clear errors — a bad link, a missing video, or disabled comments is reported with its reason in the status message and the run's OUTPUT record, never as a paid row.
• 💸 Cheap — a flat $1.20 per 1,000 comments.

🎯 Use cases

• 😀 Sentiment analysis — feed comments into your own classifier or an LLM.
• 📣 Brand and creator monitoring — see how audiences react to launches and sponsored videos.
• 🧠 Audience research — find questions, requests, and complaints in comment sections.
• 🏆 Giveaways and community work — export commenters for moderation or contests.
• 📚 Academic datasets — collect public comments for language and social research.

🚀 How to use

1. Paste one or more YouTube links into Video URLs.
2. Set Maximum comments per video (default 10).
3. Choose Newest first or Top comments, or set a Comments published after date.
4. Click Start.
5. Download the results as JSON, CSV, or Excel from the Output tab, or read them through the API.

📥 Input

{
  "startUrls": [
    { "url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ" }
  ],
  "maxComments": 10,
  "sortCommentsBy": "NEWEST_FIRST",
  "oldestCommentDate": "1 day",
  "proxyConfiguration": {
    "useScrapelyProxy": true,
    "scrapelyProxyGroups": ["DATACENTER"]
  }
}

📤 Output

One row per comment.

Example item:

{
  "cid": "UgzLAaEp8N5Q-McgwOJ4AaABAg",
  "replyToCid": null,
  "type": "comment",
  "publishedTimeText": "6 minutes ago",
  "comment": "Still a classic.",
  "author": "@janedoe",
  "authorIsChannelOwner": false,
  "replyCount": 0,
  "voteCount": 0,
  "hasCreatorHeart": false,
  "videoId": "dQw4w9WgXcQ",
  "pageUrl": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
  "commentsCount": 2459071,
  "title": "Rick Astley - Never Gonna Give You Up (Official Video) (4K Remaster)"
}

If a video cannot be read, no row is saved for it. The run's OUTPUT record lists it under failedInputs with the URL, a code (NO_RESULTS, VIDEO_UNAVAILABLE, AGE_RESTRICTED, or NO_COMMENTS) and a short reason, and the status message names it.

💰 Pricing

$0.0012 per comment ($1.20 per 1,000). Videos that cannot be read cost nothing. Platform usage (compute and datacenter proxy, billed at standard rates) is added on top and is small.

Worked example: 10 comments cost 10 × $0.0012 = $0.012 in result fees. Including platform usage, our measured cost for a typical 10-comment job is about $0.014.

❓ FAQ

Is it legal to scrape YouTube comments?

The actor reads public comments that anyone can see without logging in. Make sure your use follows YouTube's terms and privacy laws such as GDPR.

Do I need an API key, login, or proxy?

No key and no login. Datacenter proxies are used by default.

How many comments can I get?

As many as the video has. Set a high limit and the actor pages through all top-level comments.

Does it include replies?

Not as separate rows. Only top-level comments are returned, with replyCount showing how many replies each one has.

What happens if comments are turned off?

No rows are saved for that video and nothing is charged; OUTPUT.failedInputs lists it with the code NO_COMMENTS.

Can I schedule it or use the API?

Yes. Use Schedules for recurring runs, or start it through the REST API and fetch the dataset in any format.

⚠️ Limits

• Top-level comments only. Reply threads are not expanded.
• publishedTimeText is relative ("6 minutes ago"), not an exact timestamp.
• Private or deleted videos return no rows; they are listed in OUTPUT.failedInputs.
• Like counts shown in short form, such as 104K, are rounded by YouTube.
• Very short runs include a few seconds of container startup.