b站视频转学习笔记

概述

Bilibili Video to Notes

Overview

Transform a Bilibili video into structured, screenshot-rich learning notes (Markdown + HTML). The workflow downloads the video with yt-dlp, extracts keyframe screenshots using imageio-ffmpeg, inspects the frames, and generates a polished HTML report with embedded images and a matching Markdown source.

When to Use

Trigger this skill when the user:

• Pastes a Bilibili URL and asks for notes/summary/screenshots;
• Requests "学习笔记" (learning notes) from a Bilibili video;
• Wants keyframes extracted and explained from a Bilibili lecture or demo.

Workflow

1. Prepare Environment

Ensure the managed Python environment is ready:

"C:/Users/johnn/.workbuddy/binaries/python/versions/3.13.12/python.exe" -m venv "C:/Users/johnn/.workbuddy/binaries/python/envs/default"
"C:/Users/johnn/.workbuddy/binaries/python/envs/default/Scripts/pip" install --upgrade yt-dlp imageio-ffmpeg

If yt-dlp fails with HTTP 412 on Bilibili, upgrade it:

pip install --user --upgrade yt-dlp

2. Create Tasks

Create and track three tasks:

1. Download Bilibili video
2. Extract key screenshots
3. Create Markdown + HTML learning notes

3. Download Video

Get metadata first:

yt-dlp --print "%(title)s|%(duration)s|%(uploader)s|%(upload_date)s" "<URL>"

Then download the 1080p video:

mkdir -p "D:/Document/WorkBuddy/MPDV/output/video_notes_N_<topic>"
cd "D:/Document/WorkBuddy/MPDV/output/video_notes_N_<topic>"
yt-dlp -f "30080+30280" --merge-output-format mp4 -o "<base>.%(ext)s" "<URL>"

If ffmpeg is missing, the download produces a .mp4 video file and a separate .m4a audio file. The .mp4 is sufficient for screenshot extraction.

4. Extract Screenshots

Use the bundled script when possible:

/c/Users/johnn/.workbuddy/binaries/python/envs/default/Scripts/python.exe 
  C:/Users/johnn/.workbuddy/skills/bilibili-video-to-notes/scripts/download_and_extract.py 
  "<URL>" "D:/Document/WorkBuddy/MPDV/output/video_notes_N_<topic>" 
  --filename-base <base> --interval <seconds>

If the duration probe fails, use the inline hardcoded-duration approach described in references/workflow.md.

Screenshot interval guidance:

• ~30 min video: 75s interval
• ~40 min video: 90s interval
• ~50 min video: 90s interval

5. Inspect Screenshots

Use the Read tool to view screenshots in batches. Identify title slides, conceptual diagrams, UI demos, KPI tables, and summary slides. Skip duplicate frames caused by static slides.

6. Generate Notes

Create two deliverables in the output folder:

• 学习笔记_<topic>.md
• 学习笔记_<topic>.html

Recommended structure:

1. Video metadata header (title, source, duration, date);
2. Table of contents;
3. 10–15 sections following the video flow;
4. Embedded screenshots with timestamp captions;
5. Comparison/summary tables;
6. Integration opportunities (e.g., Sage X3);
7. Screenshot index appendix.

HTML style conventions:

• Light theme;
• Hero header with metadata badges;
• Card-based sections with numbered badges;
• Tables, code chips, and callouts;
• Relative screenshots/ paths.

7. Update Project Memory

Append a brief entry to:

• D:/Document/WorkBuddy/MPDV/.workbuddy/memory/YYYY-MM-DD.md
• D:/Document/WorkBuddy/MPDV/.workbuddy/memory/MEMORY.md if the content is reusable long-term.

8. Present Results

Call present_files with the HTML first and 8–12 representative screenshots.

Resources

• scripts/download_and_extract.py — Downloads a Bilibili video and extracts keyframe screenshots.
• references/workflow.md — Detailed workflow reference, troubleshooting table, and environment-specific commands.