Weibo Scraper: Weibo Search, Hot Search, Comments, Profiles
Pricing
from $2.50 / 1,000 posts
Weibo (微博) scraper, no login: weibo posts by id or link with the full long text, weibo profiles with exact followers, search by keyword or #topic#, comments and replies, likers, followers and following, the 微博热搜 hot search board with changes between runs. Optional sentiment. Misses free.
Weibo posts, one per line, in any form: the 16-digit id (mid) 5346699644242007, the 9-character id (bid) RjAS0hNxt, or a link such as https://weibo.com/1642909335/RjAS0hNxt, https://m.weibo.cn/detail/5346699644242007, https://m.weibo.cn/status/RjAS0hNxt. One row per post: the same post given several ways is one row. A value that is not a Weibo post id or link comes back as a free row that says why.
[]Weibo users, one per line: the numeric id 1642909335, a profile link https://weibo.com/u/1642909335, https://weibo.com/1642909335, https://m.weibo.cn/u/1642909335, https://weibo.com/n/微博小秘书, or a screen name 微博小秘书 / @微博小秘书. One exact profile row per user (the exact follower integer, not the rounded "1.68亿").
[]Search terms, one per line. A plain word or phrase reads page 1 of each tab in searchTabs (about 10–15 posts a tab, about 50 on 图片 pictures); #topic# reads the topic page instead. Weibo shows a visitor page 1 only and no date window: the rows say so.
[]The search tabs read for each keyword, page 1 each; empty = all (综合). A post found in two tabs is one row with both tabs in search_tabs. users gives user rows, topics gives up to 10 hot_item rows with discussion and read counts, each topic once; the other tabs give post rows.
[]The most post rows per keyword over its tabs (0 = every post of the pages read: about 90 over the five post tabs).
The hot-search boards to read, one hot_item row per entry: rank, word, heat, label (新 / 热 / 沸), category, time on board, the pinned topic and promoted entries flagged.
[]For each ranked word of the realtime board, the top posts of Weibo's 综合 search for it (one search page per word; the pinned topic and promoted entries are skipped; a post two words share is one row, and the second word takes the next post of its page; needs realtime in hotBoards); 0 = none.
How many posts of Weibo's hot (trending) timeline to deliver, 10 a page; 0 = none. The walk ends when a page adds nothing new (about 100 posts in practice).
A search, timeline or hot-timeline row cuts a long post at about 140 characters; with this on, each such post is read in full (one request per long post). Off, the row keeps the cut text and text_truncated is true.
After the posts, one exact user row per distinct author (the exact follower integer; charged as a profile). Every post row already carries the author's name, id, verification and counts as the post page shows them.
Comment rows per post row (comment: text, time, likes, region, the commenter with exact followers), to the post's own count; 0 = none. Applies to every post row of the run.
Read the replies under each comment that has them (one request per such comment), as comment rows with is_reply and parent_comment_id. Counted in maxCommentsPerPost.
Who liked the post, as user rows (list_kind liker), up to 200 a post (Weibo shows a visitor no more); 0 = none.
The newest posts of each user in users, as post rows; 0 = the profile only. Weibo shows a visitor the newest 10 and no more.
The user's followers (fans) as user rows, 10 a request, with the exact total in list_total; 0 = none. Weibo serves a visitor the first few thousand of a big account.
The accounts the user follows, as user rows, 10 a request, to the full count; 0 = none.
The user's video posts as post rows (video URL, duration, title); 0 = none.
One more request per user in users: registration date, credit level and the profile labels, and the birthday, company and school where the user shows them. Weibo shows a visitor no IP location.
Typed tags on every post and comment row from a classification model (Jev, pinned version): analysis_sentiment (−1 to 1), analysis_kind (news, opinion, promotion, entertainment, personal, complaint, question, other) and the confidence of each. Charged per tagged row.
A language code (en, de, ja, pt-BR): a machine translation of each post, comment and hot-board word by an OpenAI model, next to the original (translation, with the model named in translation_note). Empty = no translation. Charged per translated row.
Keep Weibo's own JSON object on each row (raw), for fields this Actor does not name.
The most charged rows this run delivers, all types together. 0 = no limit. Free rows (not found, invalid, empty) do not count.
http://user:pass@host:port — a proxy of yours to read Weibo through, tried before the Actor's own exits. Leave empty: the Actor brings its own proxies and no Chinese exit is needed.
A label copied onto every row (run_tag), to tell runs apart in a shared dataset.
A note to the people who maintain this Actor: what you needed and did not get. For example a missing field, a wrong or empty result, or the Actor you switch from or to, and why. The run does not read it: the rows and the charges stay the same. A person reads every note.
Comments order
commentsOrder
Optional
The order Weibo lists the comments in. Time order reaches more of a big post's comments (98 % of 1,577 in our test); hot order follows Weibo's ranking.