YouTube Comments Reliable API

Scrape public YouTube comments and replies from watch URLs, youtu.be, Shorts, Live, Embed URLs, or raw video IDs. Get clean comment text, author/channel details, like counts, reply counts, creator-heart status, relative publish time, and optional approximate timestamps without needing a YouTube Data API key.

Why this Actor?

Comment scrapers often fail mid-run when YouTube changes a continuation response or blocks a cloud IP. This Actor is built around predictable batches:

• Fast direct access is tried first.
• If YouTube blocks or returns no comments, retries automatically fall back to standard Apify Proxy.
• Proxy retries use fresh sessions.
• Duplicate video IDs and comment IDs are removed.
• Invalid or comments-off videos return uncharged diagnostic rows instead of crashing the full batch.
• Billing happens only on comment/reply rows actually written to the dataset.
• The Actor stops cleanly at the user's maximum charge limit.

Features

• Watch, youtu.be, Shorts, Live, Embed, and raw video-ID input
• Top comments or newest-first sorting
• Optional replies
• Per-video and global result limits
• Minimum-like filter
• Case-insensitive text filter
• Comment ID and parent-comment linkage
• Author name, channel ID/link, profile image
• Parsed integer like/reply counts
• Creator-heart flag
• Paid-comment text when YouTube exposes it
• Relative publish text exactly as returned by YouTube
• Optional approximate UTC timestamp clearly marked as approximate
• Direct-first access with optional automatic Apify Proxy fallback and custom-proxy support
• Retry/backoff with a new proxy session per attempt
• Structured, uncharged video-level failure records
• Pay only for successful comment/reply rows

Input example

{
  "videoUrls": [
    "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
    "https://youtube.com/shorts/VIDEO_ID"
  ],
  "sortBy": "top",
  "maxResultsPerVideo": 500,
  "maxTotalResults": 5000,
  "includeReplies": true,
  "minLikes": 0,
  "containsText": ""
}

Output

Each successful dataset row represents one public comment or reply and includes:

• videoId, canonical videoUrl
• commentId, parentCommentId, isReply
• text
• author, author channel ID/link, profile image
• normalized likeCount and replyCount
• publishedTimeText
• publishedAtApprox and publishedAtIsApproximate
• isHearted
• paidText when available

Video-level failures use recordType: "error" and are not charged.

About dates

YouTube's comment interface often returns relative time such as "3 days ago" instead of an absolute timestamp. This Actor preserves that source string. When the downloader can interpret it, publishedAtApprox is provided, but it is deliberately marked approximate rather than pretending a relative date is exact.

Filtering

minLikes and containsText are applied before output. Filtered comments are not billed. When filters are selective, the Actor may scan more source comments than it emits, with an internal safety cap to avoid unbounded work.

Reliability and proxies

The Actor uses direct-first mode because direct Apify cloud access can be much faster when YouTube accepts the request. If the first attempt is blocked or returns no comments, retries automatically switch to standard Apify Proxy. The Actor does not force residential proxy traffic.

You can disable automatic fallback or provide an explicit custom/Apify proxy configuration. Proxy retries use fresh sessions. Network requests use bounded timeouts, and continuation retries use short backoff rather than waiting indefinitely.

Pricing

Target Store price: $0.0006 per successfully emitted comment or reply — about $0.60 per 1,000 results — plus the small Actor-start event shown in the Store pricing panel.

Error records, filtered-out comments, retries, and duplicate inputs do not trigger the comment event charge.

This price sits between newer low-cost competitors around $0.40–$0.50 per 1,000 and the established high-usage Actor around $0.90 per 1,000, while adding stronger failure handling, spend controls, proxy rotation, filters, normalized engagement counts, and explicit date-quality semantics.

Limits

• Maximum 100 input videos per run
• Maximum 10,000 emitted rows per video
• Maximum 100,000 emitted rows per run
• Public comments only
• No account/login bypass
• Deleted/private content is not bypassed

Use scraped public data in accordance with applicable platform terms, copyright, privacy, and data-protection rules.

Support

For a reproducible issue, include the public video URL, sort mode, proxy setting, and Apify run ID. Never post API tokens or private proxy credentials.