Xiaohongshu Video Downloader & Summarizer

Download videos from Xiaohongshu (小红书/RedNote) and optionally generate a full resource pack: video + audio + subtitles + transcript + AI summary.

Workflow

Follow these 6 steps in order:

Step 1: Check Dependencies

Verify required tools are installed:

yt-dlp --version
ffmpeg -version

Both must be available. If missing, instruct the user to install:

• brew install yt-dlp (macOS) or pip install yt-dlp
• brew install ffmpeg (macOS)

Step 2: Get Video Information

/opt/homebrew/Caskroom/miniconda/base/envs/myenv/bin/python scripts/download_xiaohongshu.py "URL" --list-formats

This shows available formats and verifies the URL works with cookie authentication.

Step 3: Download Video + Extract Audio

For basic download (video only, backward compatible with v1.0):

/opt/homebrew/Caskroom/miniconda/base/envs/myenv/bin/python scripts/download_xiaohongshu.py "URL"

For full resource pack (video + audio + subtitles + transcript):

/opt/homebrew/Caskroom/miniconda/base/envs/myenv/bin/python scripts/download_xiaohongshu.py "URL" --full

For full resource pack + AI summary preparation:

/opt/homebrew/Caskroom/miniconda/base/envs/myenv/bin/python scripts/download_xiaohongshu.py "URL" --summary

The --full flag creates a folder ~/Downloads/<video title>/ containing:

• video.mp4 — original video
• audio.mp3 — extracted audio
• subtitle.vtt — WebVTT subtitles (via 3-tier strategy)
• transcript.txt — plain text transcription

The --summary flag implies --full and additionally saves .meta.json for AI summary generation.

Step 4: Subtitle Acquisition (Automatic — 3-Tier Strategy)

The script automatically tries these strategies in order:

1. Manual subtitles — yt-dlp --write-subs --sub-lang zh,en,zh-Hans,zh-CN
2. Auto-generated subtitles — yt-dlp --write-auto-subs
3. Whisper local transcription — Falls back to parallel_transcribe.py using faster-whisper

Step 5: Generate Transcript (Automatic)

The script automatically strips timestamps from VTT to produce transcript.txt.

Step 6: AI Summary (Claude generates summary.md)

If the user requested a summary (via --summary flag or by asking to "summarize"):

1. Read the transcript file:

   ~/Downloads/<video title>/transcript.txt
   

2. Read the metadata file:

   ~/Downloads/<video title>/.meta.json
   

3. Read the summary prompt template:

   reference/summary-prompt.md
   

4. Replace the template placeholders with actual values:

   • {{TITLE}} — from .meta.json
   • {{URL}} — from .meta.json
   • {{DURATION}} — from .meta.json
   • {{PLATFORM}} — "Xiaohongshu (小红书)"
   • {{TRANSCRIPT}} — contents of transcript.txt

5. Generate the summary following the template structure.

6. Save the result to:

   ~/Downloads/<video title>/summary.md
   

Options Reference

Output Structure

Basic mode (default)

~/Downloads/
└── <title> [<id>].mp4

Full resource pack mode (--full or --summary)

~/Downloads/<video title>/
├── video.mp4          # Original video
├── audio.mp3          # Extracted audio
├── subtitle.vtt       # WebVTT subtitles
├── transcript.txt     # Plain text transcript
├── .meta.json         # Video metadata (--summary only)
└── summary.md         # AI-generated summary (--summary only, written by Claude)

Supported URL Formats

Troubleshooting

Important Notes

• Always use the full share URL (with xsec_token) for best results
• Log into xiaohongshu.com in your browser before downloading
• Maximum video quality is typically 1080p (platform limitation)
• Whisper transcription requires uv for automatic dependency management, or faster-whisper installed manually
• Respect copyright and Xiaohongshu's terms of service