About this skill

What problem it addresses

Bilibili danmaku is not just a comment feed; timing, density, and recurring keywords often reveal audience sentiment better than the video transcript alone. Manually counting danmaku, screenshotting samples, and tallying positive or negative phrases is slow and hard to reproduce. This skill turns a video link into structured evidence: it extracts danmaku, splits it by timeline, computes keyword shares, and produces an LLM prompt plus a Markdown report.

How it works

It uses public Bilibili APIs to fetch video metadata and danmaku without login. The main steps are:
- Extraction: pull rolling, bottom, and top danmaku into analyzable text.
- Timeline segmentation: split the video into 10 segments to inspect pacing and density shifts.
- Sentiment statistics: calculate positive, negative, and question keyword shares using built-in lists such as positive keywords, negative keywords, and question keywords.
- Sampling: keep representative samples per segment to control prompt length.
- Output: generate JSON data, an analysis prompt file, and a Markdown report.

The report usually includes video info, danmaku statistics, time distribution, segment samples, sentiment analysis, pacing analysis, risk notes, and a summary. It is not intended to replace the LLM’s judgment; it prepares auditable, reusable input for deeper interpretation.

Limits

It fits public Bilibili videos and is useful for sentiment, pacing, and interaction analysis. API rate limits can return 412 or 403; empty danmaku may require checking the CID, network, or video state. Keyword counts are heuristic, so slang, sarcasm, and context still need the LLM to interpret the sampled segments.

Screenshots

Use Cases

After a video ships, quickly identify the timestamp with the highest danmaku density to spot audience reaction points.
During an ops review, summarize positive, negative, and question danmaku shares into a sentiment summary.
When a hot video goes live, extract segmented danmaku samples to check controversial memes and negative signals.
For content research, compare danmaku pacing across videos and compile time distribution and peak density segments for a report.

Best For

Content operations owners reviewing Bilibili videos need to turn danmaku density and sentiment shares into report-ready summaries.
Product or community managers monitoring public sentiment need to spot negative danmaku and controversial memes after a hot video launch.
Short-video or stream-clip researchers need to compare viewer interaction pacing across time segments.
Community risk maintainers need to collect representative danmaku samples and evidence for risk notes.