YouTube Comments Scraper

---

🎯 What this scrapes

YouTube's own video page walks a two-step internal API to load comments: fetch the watch payload, pull a continuation token out of its comments-section engagement panel, then page through /youtubei/v1/next for the actual comment nodes. We do the same walk — a generic recursive search finds that token wherever YouTube's payload nests it this month, so a client-side layout shift doesn't quietly break the scrape — and turn every comment and reply into one clean dataset row, sorted by top or newest comments, with optional full reply threads.

🔥 What we handle for you

• 🛡️ Browser fingerprint rotation — curl-cffi impersonates real Chrome / Firefox / Safari TLS handshakes so the target sees a browser, not Python.
• 🌐 Residential proxy rotation via Apify Proxy — fresh session and exit IP on every block.
• 🔁 Retries with exponential backoff on 408 / 429 / 5xx — up to 5 attempts per page, Retry-After honoured.
• 🧱 Rate-limit-aware pacing — when the target pushes back, we slow down instead of getting banned.
• 🧊 Clean, typed dataset rows — Pydantic-validated, ISO-8601 timestamps, stable IDs, JSON / CSV / Excel export straight from the Apify Console.
• 💰 Pay-Per-Event pricing — you only pay for results that hit your dataset. No data, no charge.

💡 Use cases

• Social listening — pull every comment on a video the moment it lands and feed sentiment/topic models in near real time.
• Brand monitoring — track what viewers say under a brand's own uploads or under mentions in a competitor's videos.
• Creator analytics — give creators and their teams a clean export of audience reaction, sorted top or newest, without babysitting the YouTube UI.
• NLP pipelines — bulk comment/reply text with stable IDs for training or evaluating classifiers, topic models, and toxicity filters.

⚙️ How to use it

1. Click Try for free at the top of the page.
2. Fill in the input form — most fields have sensible defaults.
3. Click Start. Output streams into the run's dataset.
4. Export from Storage → Dataset as JSON, CSV, or Excel — or fetch via the API.

📥 Input

Field	Type	Required	Default	Notes
videoUrls	array	yes	['https://www.youtube.com/watch?v=dQw4w9WgXcQ']	YouTube video URLs, short URLs (youtu.be/...), Shorts URLs, or bare 11-character video IDs. Mix and match…
maxComments	integer	no	100	Stop paging once this many top-level comments (and replies, if included) have been emitted for a single video…
sortBy	string	no	'top'	Comment order YouTube returns — matches the sort control on the video page itself.
includeReplies	boolean	no	False	Follow each top-level comment's reply thread too. Multiplies request volume and cost — leave off unless you need full…
hl	string	no	'en'	YouTube UI language passthrough — affects relative-timestamp text like "1 year ago". Example: en.
gl	string	no	'US'	YouTube UI country/region passthrough. Example: US.
proxyConfiguration	object	no	{'useApifyProxy': True}	Apify Proxy configuration. Recon confirmed byte-identical comment payloads across datacenter and residential exits —…

Example input

{
  "videoUrls": [
    "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
    "https://www.youtube.com/watch?v=jNQXAC9IVRw"
  ],
  "maxComments": 45,
  "sortBy": "top",
  "includeReplies": false,
  "hl": "en",
  "gl": "US",
  "proxyConfiguration": {
    "useApifyProxy": false
  }
}

📤 Output

Every row is one dataset item.

Field	Type	Notes
video_id	string	11-character YouTube video ID this comment belongs to.
comment_id	string	YouTube's opaque comment/reply ID.
author	string	Commenter's display name at scrape time.
author_channel_id	string	Commenter's channel ID. Null when YouTube omits it (e.g. deleted account).
text	string	Comment or reply body, plain text.
published_time	string	YouTube's relative publish string (e.g. '1 year ago'), verbatim — YouTube does not expose an absolute timestamp on this…
like_count	integer	Like count at scrape time. Defaults to 0 when absent.
reply_count	integer	Number of replies under this top-level comment. 0 for replies themselves.
is_pinned	boolean	True when the video owner pinned this comment.
is_hearted	boolean	True when the video owner hearted this comment.
is_reply	boolean	True for a reply row, false for a top-level comment.
parent_comment_id	string	The top-level comment_id this row replies to. Null for top-level comments.

Example output

{
  "video_id": "dQw4w9WgXcQ",
  "comment_id": "Ugzge340dBgB75hWBm54AaABAg",
  "author": "SomeUser",
  "author_channel_id": "UCxxxxxxxxxxxxxxxxxxxxxx",
  "text": "can confirm: he never gave us up",
  "published_time": "1 year ago",
  "like_count": 1204,
  "reply_count": 3,
  "is_pinned": false,
  "is_hearted": false,
  "is_reply": false,
  "parent_comment_id": null
}

💰 Pricing

Pay-Per-Event — you pay only when these events fire:

Event	USD	What it is
actor-start	$0.2	One-off warm-up charge per run
comment-scraped	$0.0005	PPE event

Example: 1 000 results at the rates above ≈ $0.70. No subscription, no minimum, no card to start — Apify gives every new account $5 of free credit.

🚧 Limitations

Public comments only — no OAuth, no moderator tools, no comment-author metadata beyond what YouTube's own comments UI exposes. YouTube's internal payload shape can shift between web-client releases; we walk it with a generic recursive search rather than a fixed path, but a structural change may still need a same-day patch. Enabling replies multiplies request volume per video — factor that into cost expectations for reply-heavy videos.

❓ FAQ

Does this scrape live chat?

No — this is the standard comments section under a video, not live-stream chat. For live chat you want a different endpoint entirely.

Why are replies not included by default?

Following every reply thread multiplies request volume (and cost) per video. Turn on includeReplies when you need full threads; leave it off for a fast top-level-only pull.

Why is published_time a string like '1 year ago' instead of a date?

That's what YouTube's own comments endpoint returns — it does not expose an absolute timestamp here. We pass it through verbatim rather than guessing at a conversion.

Why did one video return zero comments?

Comments can be disabled by the uploader, or the video may have none yet. The run still succeeds — it reports how many videos had no comments in the final status message rather than failing the whole run.

What happens if a video ID is invalid or private?

That single video is skipped and counted as failed in the run summary; every other video in the same run still completes.

💬 Your feedback

Spotted a bug, hit a weird edge case, or need a new field? Open an issue on the Actor's Issues tab on Apify Console — we ship fixes weekly and we read every report.

---