Douyin Comment Tracker

Watch a set of Douyin (抖音) videos and pull their comments on a schedule — then export flat, CSV-ready rows: comment text, author, likes, reply count, IP region, timestamps, stickers. Douyin is the mainland-China app TikTok is based on. This Actor calls Douyin's web API directly with signed HTTP requests. No headless browser, so it is faster, cheaper, and steadier under load than browser-driven scrapers.

Built for ongoing social listening on Douyin: brand and campaign monitoring, KOL / influencer tracking, sentiment analysis, and competitive research. Run it once for a bulk export, or point a schedule at it and let delta mode feed you only what's new.

Why this one

Most Douyin comment scrapers do a single bulk dump. This Actor is built to be run again and again on the same videos without re-paying for or re-processing comments you already have:

• Delta mode — each scheduled run returns only comments new since the previous run (per video), tracked by a high-water mark in the key-value store. A daily "watch these 50 videos" job stays cheap and its dataset stays clean.
• Standby / HTTP endpoint — call the Actor as a low-latency GET /comments?videoUrl=… service for one-off lookups, no run to start.
• Flat rows, one dataset — every row carries a row_type; no nested JSON, opens straight in Excel / Sheets / a CSV import.
• Per-video error isolation — a private / deleted / comments-off video produces one typed error row and the run keeps going. You are never charged for error rows, empty results, or delta duplicates.

What you get

Without any login — the common case:

• Top-level comments for any public video, with pagination handled for you: comment_id, text, create_time (ISO + epoch), digg_count (likes), reply_count, user_id, sec_uid, nickname, avatar_url, ip_label (the region Douyin shows), is_author_liked, sticker_url.
• Douyin video URLs, v.douyin.com share links, and bare aweme_id all accepted; share links are resolved for you.
• A typed error row per video that can't be scraped (private, deleted, comments off, region-locked) — one bad video never fails the run.

With a cookies string from a logged-in douyin.com session:

• Nested reply threads under each comment (row_type: "reply", linked by parent_comment_id).
• Video metadata rows (row_type: "video") — description, author, play / like / comment / share counts, music.

Douyin gates its reply and video-detail endpoints behind a logged-in session, so these two need cookies. Without them the Actor scrapes top-level comments and logs that it skipped the rest — it does not fail. See Limitations below.

Input

Field	Type	Default	Notes
videoUrls	string[]	— (required)	Douyin video URLs, v.douyin.com share links, or bare numeric aweme_id.
maxCommentsPerVideo	integer	100	Top-level comments per video. Replies don't count against this. Capped at 5000.
includeReplies	boolean	false	Fetch nested reply threads. Needs cookies — skipped without it.
maxRepliesPerComment	integer	10	Replies fetched per comment when includeReplies is on.
includeVideoMetadata	boolean	true	Emit one row_type: "video" row per video. Works best with cookies.
deltaMode	boolean	false	Only return comments new since the previous run.
deltaStateKey	string	DOUYIN_DELTA_STATE	Key-value store key holding delta state. Reuse it across scheduled runs of one job.
cookies	string	—	name=value; name=value cookie string from a logged-in douyin.com session. Unlocks replies + metadata. Do not share this value.
proxyConfiguration	object	RESIDENTIAL Apify Proxy	Datacenter IPs are blocked quickly — keep residential.
maxConcurrency	integer	3	Videos in parallel, and the per-host request cap. Capped at 10.

Example input

{
  "videoUrls": [
    "https://www.douyin.com/video/7677636831337576185",
    "https://v.douyin.com/iRxYpVKN/",
    "7677636831337576185"
  ],
  "maxCommentsPerVideo": 200,
  "includeVideoMetadata": true,
  "proxyConfiguration": { "useApifyProxy": true, "apifyProxyGroups": ["RESIDENTIAL"] },
  "maxConcurrency": 3
}

Output

One flat row per comment. Every row carries a row_type (comment, reply, video, or error) so you keep everything in one dataset and filter after export. No nested objects — it opens cleanly in Excel / Google Sheets / a CSV import.

Comment row

{
  "row_type": "comment",
  "source": "https://www.douyin.com/video/7677636831337576185",
  "aweme_id": "7677636831337576185",
  "comment_id": "7355800000000000001",
  "parent_comment_id": null,
  "text": "这个视频太有意思了 😂",
  "create_time": "2024-05-01T09:12:30+00:00",
  "create_time_epoch": 1714554750,
  "digg_count": 128,
  "reply_count": 4,
  "user_id": "58312900000",
  "sec_uid": "MS4wLjABAAAA…",
  "nickname": "小明同学",
  "avatar_url": "https://p3-pc.douyinpic.com/…/avatar.jpeg",
  "ip_label": "广东",
  "is_author_liked": true,
  "is_author_reply": false,
  "sticker_url": null,
  "reply_to_reply_id": null,
  "scraped_at": "2026-09-02T12:00:00+00:00"
}

A reply row (needs cookies) is the same shape with row_type: "reply" and parent_comment_id set to the top-level comment's id.

Video-metadata row (needs cookies)

{
  "row_type": "video",
  "aweme_id": "7677636831337576185",
  "desc": "周末去了趟海边 #vlog",
  "create_time": "2024-05-01T08:00:00+00:00",
  "create_time_epoch": 1714550400,
  "share_url": "https://www.douyin.com/video/7677636831337576185",
  "duration_ms": 34000,
  "author_user_id": "58312900000",
  "author_sec_uid": "MS4wLjABAAAA…",
  "author_nickname": "旅行的阿May",
  "author_follower_count": 254000,
  "statistics_play_count": 1830000,
  "statistics_digg_count": 96000,
  "statistics_comment_count": 1204,
  "statistics_share_count": 3100,
  "statistics_collect_count": 8800,
  "music_id": "6935000000000000000",
  "music_title": "原声 - 旅行的阿May",
  "music_author": "旅行的阿May"
}

Error row

{
  "row_type": "error",
  "source": "https://www.douyin.com/video/0000000000000000000",
  "aweme_id": "0000000000000000000",
  "error_code": "VIDEO_NOT_FOUND",
  "error_message": "…",
  "http_status": null,
  "scraped_at": "2026-09-02T12:00:00+00:00"
}

Error codes: SIGNATURE_REJECTED, RATE_LIMITED, VIDEO_NOT_FOUND, VIDEO_PRIVATE, COMMENTS_DISABLED, PROXY_FAILED, PARSE_ERROR, INPUT_ERROR.

Pricing (pay-per-event)

You are billed per result actually written to the dataset:

Event	Fires when
Actor start	Once per run.
Comment	One top-level comment row is written.
Reply	One nested reply row is written.
Video metadata	One video-metadata row is written.

You are never charged for an error row, for an empty result, or — in delta mode — for a comment you already received on a previous run. Set a maximum cost per run to cap spend; the Actor stops cleanly at the limit and keeps everything already produced and billed.

Delta mode is the cheapest way to monitor a set of videos: after the first run, each scheduled run only pays for genuinely new comments.

Standby mode

Enable Standby to call the Actor as a low-latency HTTP endpoint instead of starting a run:

GET /comments?videoUrl=<url|aweme_id>&maxComments=50

It returns JSON directly, bounded to ~120 comments / a few seconds per request — larger jobs get a clear "use batch mode" error. Standby and batch share the same scraping core, so results match.

Limitations — read this

• Top-level comments work without login. Replies and video metadata need the cookies input — Douyin gates its reply and video-detail endpoints behind a logged-in session (a device uifid token plus, for metadata, an extra request signature). Without cookies the Actor scrapes top-level comments and logs a skip for the rest; it does not fail.
• Private / friends-only videos cannot be scraped — you get a VIDEO_PRIVATE / VIDEO_NOT_FOUND error row.
• Comments turned off returns COMMENTS_DISABLED. Douyin does not always distinguish "comments off" from "zero comments yet", so a brand-new video can also land here.
• Region-restricted content: some videos are only visible from certain regions. Residential-proxy country selection can help; otherwise you get an error row.
• Rate limits: Douyin rate-limits aggressively. Residential proxies are effectively required — datacenter IPs are blocked fast. Large jobs run slower because the Actor paces itself and rotates proxy sessions to stay under the radar.
• Comment ordering is Douyin's own (a mix of "hot" and recent), not strictly chronological. Delta mode tracks the highest comment id seen rather than assuming order.
• No historical backfill guarantee: Douyin's comment API stops paginating at some depth for very high-comment videos.
• Scrapes public data only and does not log in on your behalf. Supplying cookies is optional and at your own risk. Respect Douyin's terms and applicable law; you are responsible for how you use the data.

How it works

The Actor generates Douyin's web request signature (a_bogus) in-process, attaches the anti-abuse ttwid cookie and an msToken, and calls the same https://www.douyin.com/aweme/v1/web/comment/list/ endpoint the Douyin website uses. The signature layer is isolated in one module so it can be updated fast when Douyin rotates the algorithm. See MAINTENANCE.md.

---

Douyin, 抖音, Douyin comments, Douyin comment scraper, Douyin comment tracker, Douyin comment monitor, 抖音评论, 抖音评论采集, 抖音评论监控, 抖音数据采集, TikTok China, TikTok China comments, douyin.com scraper, short-video comments, 短视频评论, social listening, 舆情监测, brand monitoring, campaign monitoring, KOL research, 达人分析, influencer research, sentiment analysis, 情感分析, audience insights, comment export, CSV, scheduled scraping, delta scraping.