Wechat Link To MP4 — Development skill for Claude Code

wechat-link-to-mp4

English | [简体中文](README.zh-CN.md)

Turn a **WeChat Channels (视频号) share link** into a local mp4, plus a **timestamped transcript** and **contact-sheet thumbnails** that are easy to hand to an AI for analysis. Works as a command-line tool or as a Claude Code skill (see `SKILL.md`).

Windows + WeChat for Windows (4.x) only. Only download content you have the right to save, for personal use.

How it works

1. Starts a temporary mitmproxy instance and sets it as the system proxy. Only qq.com traffic is intercepted; everything else is passed through untouched. It can chain to an existing upstream proxy.
2. Drives the WeChat desktop window: searches for 文件传输助手 (File Transfer), verifies the chat title before sending anything, sends the link, then clicks it.
3. The proxy injects a small snippet into the Channels page JS to capture the video URL, urlToken and decodeKey.
4. Downloads the video and decrypts its first 128 KB with an ISAAC64 keystream (the scheme Channels uses), then checks the mp4 header.
5. Restores your original proxy settings, whether the run succeeds or fails.
6. Grabs a frame every N seconds and tiles them into 3×2 contact sheets; transcribes speech with faster-whisper.

Install

pip install -r requirements.txt

• Dependencies: mitmproxy requests opencv-python numpy pillow faster-whisper "av<16" opencc-python-reimplemented
• mitmproxy can also live in its own venv at ~/.wechat-link-to-mp4/venv; the script prefers it when present.
• The first run shows a UAC prompt to install the mitmproxy root certificate into Trusted Root Certification Authorities (needed to decrypt HTTPS). To remove it later, delete the mitmproxy certificate in certmgr.msc.
• The speech model (faster-whisper small, ~460 MB) downloads automatically on first use.

Usage

python scripts/wechat_link_to_mp4.py "https://weixin.qq.com/sph/xxxxxxxx"

Open and log in to WeChat for Windows