How do I export Bilibili comments in bulk? (2026)
1,000 comments come to about $0.10: page through bilibili.videos.comments.list with the video's BV number, 20 comments a page at $0.0020, and write them to a table. Replies are a separate call. There is no endpoint for danmaku, the comments shown over the video: you get only how many there are. DataXAPI reads Bilibili with 9 endpoints, from $0.0020 a call; a failed call is not charged.
Updated 2026-11-04
How to do it
One video, up to 200 comments: dataxapi run comments platform=bilibili id=BV_NUMBER pages=10 --out comments.csv writes a table of comment ID, text, commenter, likes, replies and time, for $0.02. A run takes up to 10 pages.
More than that: dataxapi call bilibili.videos.comments.list id=BV_NUMBER --pages 50 --max-spend 0.1 --out comments.json takes about 1,000 for $0.10, as JSON, which a few lines of code or your AI turn into a CSV or an Excel file. In code, pass page.next_cursor back as cursor until page.has_more is false, within the hour a cursor is good for.
The comments endpoint takes the BV number, BV1 and 12 characters in all, which is in the video's address. Only have a b23.tv short link? bilibili.videos.get takes the link and gives the number, at $0.0020.
Replies, and many videos
Comments come most liked first. Replies need the comment's id and its post_id, both in the comment list. To find the videos, use bilibili.users.videos.list for an uploader's, about 30 a page, or bilibili.videos.search for a topic's, about 20 a page.
| Job | Calls | Comes to |
|---|---|---|
| 1,000 comments of one video | 50 × $0.0020 | $0.10 |
| One page of replies under 50 comments | 50 × $0.0020 | $0.10 |
| An uploader's 10 videos, 5 pages of comments each | 1 × $0.0020 + 50 × $0.0020 | $0.102 |
Danmaku
DataXAPI has no endpoint for the danmaku themselves, so their text cannot be read here. The count can: with view=full, bilibili.users.videos.list gives each video's in ext.danmaku, and so does bilibili.videos.get.
Good to know
To go from the table to what people want or complain about, ask your AI to group the comments, with how many times each comes up and a line in their own words. Sort by likes first. Leave commenters' names out of what you share. DataXAPI reads public data only.
Endpoints
Each with its price; every page has the parameters, the fields and a recorded response.