bilibili-analyzer_skill

Overview

This skill automates analysis of Bilibili videos: it downloads a video, extracts frames, and uses AI to produce a structured topic document or hands-on tutorial. The output is a high-quality Markdown report with accurately matched screenshots and extracted code or text from frames. It focuses on reorganizing content into readable chapters rather than a raw timeline dump.

How this skill works

Given a Bilibili URL the tool downloads the video (via Bilibili APIs) and calls FFmpeg to extract frames at a configurable FPS. Frames are de-duplicated by adjacent-frame similarity and renumbered. Images are analyzed in batches by an AI worker to transcribe text, detect UI elements, and identify key steps. Finally the results are reassembled into a structured Markdown document with verified image insertions and code provenance notes.

When to use it

• You want a usable tutorial from a software demo or configuration video.
• You need a structured topic document from a lecture or knowledge talk.
• You must extract code, commands, or UI steps directly from video frames.
• You want screenshots accurately matched to sections in a generated guide.
• You need to convert long videos into summarized, actionable documentation.

Best practices

• Choose FPS based on video length (short: 1fps, medium: 0.5fps, long: 0.2fps) to balance coverage and volume.
• Keep default similarity threshold (0.80) unless many nearly-identical frames remain.
• Use --video-only or --frames-only modes when you already have partial artifacts.
• Verify key screenshots manually before publishing to ensure image-content alignment.
• Label and consolidate related frames into thematic sections instead of relying on timeline order.

Example use cases

• Convert a programming tutorial video into a step-by-step Markdown tutorial with code blocks and source frame references.
• Analyze a software walkthrough to extract installation and configuration steps with annotated screenshots.
• Summarize a lecture into a structured topic document with images illustrating key diagrams or slides.
• Process a long conference talk with reduced FPS to create an executive summary and resource list.
• Batch-analyze multiple short videos to build a curated knowledge base with consistent formatting.