Douban Scraper: Movies, Books, Ratings, Comments & Reviews (豆瓣)

Scrape Douban (豆瓣), China's biggest movie, TV and book rating site, with no login and no cookie from you. Search titles, get full details and ratings, export short comments (短评) and reviews (影评/书评) with full text, and pull charts such as "now in theaters".

$3 per 1,000 rows. That's 10× cheaper than the $30 per 1,000 charged for Douban reviews elsewhere.

中文简介：豆瓣数据采集，无需登录。支持电影/电视剧/图书搜索、条目详情（评分、评分人数、导演演员/作者出版社、简介、标签）、短评（最多约 150–200 条热门）、影评/书评（最多 40 篇热门，可选全文）、豆瓣榜单（正在热映、实时热门、热门剧集、虚构类图书等）。每 1000 条 3 美元。

Modes

Mode	Input	Rows
search	keywords, searchType movie (includes TV) or book	one row per title; fetchDetails adds full details
subject	Douban links or movie:<id> / tv:<id> / book:<id>	full details: rating and vote count, intro, directors, actors, genres, countries, release dates, durations, or authors, publisher, pages, price
comments	the same	short comments, most helpful first (up to about 150–200 per title)
reviews	the same	reviews, most helpful first (top 40 per title), fullReviewText for the whole review
chart	collection ids (movie_showing, movie_real_time_hotest, tv_hot, book_fiction…)	ranked titles

Pricing (pay per event)

Event	Price
Run start	$0.001
Row (title, comment or review)	$0.003 ($3 per 1,000)

Error rows are free. maxChargeUsd (default $5) stops the run cleanly before it goes over.

Input examples

{ "mode": "search", "searchQueries": ["三体", "活着"], "searchType": "book", "fetchDetails": true }

{ "mode": "reviews", "subjectUrls": ["https://movie.douban.com/subject/1292052/"], "maxReviews": 40, "fullReviewText": true }

{ "mode": "comments", "subjectUrls": ["https://movie.douban.com/subject/26647087/", "book:1007305"], "maxComments": 200 }

{ "mode": "chart", "charts": ["movie_showing", "tv_hot"], "maxResults": 50 }

Output examples (real rows, 2026-10-18, long text shortened)

Book (search + details):

{ "type": "subject", "detail": "full", "subjectType": "book", "id": "2567698", "title": "三体", "rating": 8.9, "ratingCount": 518798,
  "commentCount": 125038, "reviewCount": 4801, "intro": "文化大革命如火如荼进行的同时。军方探寻外星文明的绝秘计划“红岸工程”取得了突破性进展。…",
  "url": "https://book.douban.com/subject/2567698/", "subtitle": "刘慈欣 / 2008 / 重庆出版社", "authors": ["刘慈欣"], "publisher": "重庆出版社",
  "pubdate": "2008-1", "pages": "302", "price": "23.00", "bookSubtitle": "“地球往事”三部曲之一", "searchQuery": "三体", "position": 1 }

Short comment:

{ "type": "comment", "subjectType": "book", "subjectId": "1007305", "commentId": "532193706",
  "text": "盛衰之理，本为天命。然而人心就是如此。眼见得他起高楼，于是便不忍心见他楼塌了。…", "stars": 5, "votes": 11058, "status": "done",
  "createdAt": "2012-05-15T09:26:24+00:00", "authorName": "朱红尽颓", "authorId": "49117267" }

Review (with full text):

{ "type": "review", "subjectType": "movie", "subjectId": "1292052", "reviewId": "1000369", "title": "十年·肖申克的救赎",
  "stars": 5, "usefulCount": 18956, "uselessCount": 735, "commentsCount": 1082, "createdAt": "2005-05-12T12:44:13+00:00",
  "url": "https://movie.douban.com/review/1000369/", "authorName": "大头绿豆", "authorId": "1000152",
  "content": "距离斯蒂芬·金（Stephen King）和德拉邦特（Frank Darabont）们缔造这部伟大的作品已经有十年了。…", "spoiler": false }

Chart:

{ "type": "subject", "detail": "chart", "subjectType": "movie", "id": "37068446", "title": "生化危机：爆发夜", "year": "2026", "rating": 7.3,
  "ratingCount": 179182, "subtitle": "2026 / 德国 法国 美国 / 科幻 惊悚 恐怖 / 扎克·克雷格 / 奥斯汀·艾布拉姆斯 扎克·切利",
  "url": "https://movie.douban.com/subject/37068446/", "chart": "movie_showing", "rank": 1 }

Compact output for AI agents

Set Output format to compact and the run still pushes the same rows to the dataset at the same price — nothing about billing changes. It additionally writes OUTPUT_COMPACT to the run's key-value store: a lossless columnar JSON encoding of the same rows (one column-name header plus one array per row, instead of repeating every key name on every row), which is cheaper to hand to an LLM than the plain dataset JSON.

Measured on a real 150-row export of this actor's short-comment rows (see benchmarks/compact-output/ in this repo for methodology and all 5 actors' numbers):

Tokenizer	Pretty JSON	Minified JSON	Compact	vs Pretty	vs Minified
cl100k_base (GPT-4/3.5)	30,258	24,324	19,560	35.4% smaller	19.6% smaller
o200k_base (GPT-4o)	26,632	20,696	16,091	39.6% smaller	22.3% smaller

Decoding it back to the normal row objects is five lines of Python:

def decode_columnar(payload):
    cols, marker = payload["columns"], payload["missing"]
    return [{c: v for c, v in zip(cols, row) if v != marker}
            for row in payload["rows"]]

Good to know

• No-login limits: about the top 150–200 short comments and the top 40 reviews per title, sorted by helpfulness. "Latest first" ordering needs login.
• TV series links on movie.douban.com work; the actor detects TV automatically.
• Times are converted from Beijing time to UTC.
• Privacy: comment and review rows include only the author's public nickname and id. This actor does not collect avatars, locations or registration dates.

Reliability

A self-test runs every day against all endpoints. If Douban blocks every request, the run fails instead of ending green with 0 rows.

关键词： 豆瓣爬虫, 豆瓣电影数据, 豆瓣读书, 豆瓣评分, 豆瓣短评采集, 豆瓣影评, Douban scraper, Douban API, Douban movie ratings, Douban book reviews.