Bilibili Video Notes

What Bilibili Video Notes does

Bilibili Video Notes automates note-taking for Bilibili lecture videos, aimed at students taking online courses. The pipeline downloads the video and its AI subtitles, extracts a frame every 10 seconds, removes duplicates with OCR and perceptual hashing, and has a vision model score the remaining frames. The README's example reduces 129 frames to 20 to 40 and then to 7 to 12 selected screenshots for a 21-minute video. Content shown in the images is extracted and fused with the subtitle text into a structured DOCX note, and temporary files are cleaned up afterward.

It can be called directly by Hermes Agent, Claude Code, Codex CLI and other agents, and works with any OpenAI-compatible multimodal vision API, including OpenAI, Sophnet and a local vLLM server. Setup uses a .env file with VISION_API_KEY, VISION_BASE_URL and VISION_MODEL, plus a Bilibili SESSDATA cookie, because AI subtitles need a logged-in session. The README suggests pasting its install prompt to your AI and letting it do the setup.

Key features

• Downloads Bilibili video and AI subtitles
• Frame extraction every 10 seconds with OCR and perceptual-hash deduplication
• Vision model scoring that picks the most complete frame for each knowledge point
• Structured DOCX notes that merge subtitles and screenshot content
• Works with any OpenAI-compatible multimodal API, including local vLLM
• Automatic cleanup of temporary files

When to use it

• Generating study notes from a long online course lecture
• Capturing the key slides and diagrams of a video without screenshotting by hand
• Processing a local video file into a structured note

Who it is for: Students and learners who take Bilibili video courses and want agent-generated notes with screenshots.

How it fits with Hermes Agent

The repository is tagged hermes-agent and the README says Hermes Agent, Claude Code and Codex CLI can call it directly, so it supports Hermes among several agents.

How to install Bilibili Video Notes

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

git clone https://github.com/asdhabdua/bilibili-video-notes-skill.git
cd bilibili-video-notes-skill
pip install -r scripts/requirements.txt

Requirements: FFmpeg available on PATH, a multimodal vision API key set in .env, and a Bilibili SESSDATA cookie in bilibili_cookies.txt that expires after a few days.

Note: Subtitle download needs a logged-in Bilibili cookie that expires after a few days, and a multimodal vision API key is required.

FAQ