🎬 Xiaoer VideoLab

One click. Any video. Local.

Press one toolbar button and the video on the current page lands in your ~/Downloads. Powered by a tiny local yt-dlp daemon — 1800+ sites out of the box (YouTube · Bilibili · X/Twitter · TikTok · Vimeo · Twitch · Weibo …).

 

---

🙏 Thanks to our contributors / 致谢贡献者

This project is better because of these people — thank you! / 本项目因他们而更好,衷心感谢!

• @ttmouse — popup history panel, cancel a stuck download, play / open-folder, one-click daemon start (#4)
  弹出历史面板、取消卡住的下载、播放 / 打开文件夹、一键启动服务
• @jzq1212 (林以恒) — Windows support: cross-platform daemon + PowerShell install scripts (#1)
  Windows 支持:跨平台 daemon + PowerShell 安装脚本
• @alick-zhang — raised the Windows request that kicked it off (#3)
  提出 Windows 需求,促成了上面的 Windows 支持

Issues & PRs welcome — your name goes here too. / 欢迎提 issue 和 PR,你的名字也会出现在这里。

---

Why

Browser video downloaders are a swamp of sketchy extensions that beg for "read everything on every site" permissions and phone home. Xiaoer VideoLab takes the opposite bet:

• The extension does almost nothing. It only reads the current tab's URL when you click it, and POSTs that one string to 127.0.0.1. No page scraping, no content scripts, no remote servers.
• The download happens locally. A small Python daemon hands the URL to yt-dlp, the battle-tested open-source downloader. All the smarts live in a tool you can audit.
• Nothing leaves your machine except the request yt-dlp makes to fetch the video you asked for.

How it works

 ┌─────────────────────┐   click    ┌──────────────────────────┐         ┌──────────┐
 │  Browser toolbar     │ ─────────► │  daemon @ 127.0.0.1:7788 │ ──────► │  yt-dlp  │ ──► ~/Downloads
 │  (Chrome MV3 ext.)   │  POST url  │  (Python stdlib, launchd)│  spawn  └──────────┘        │
 └─────────────────────┘            └──────────────────────────┘                              ▼
        ▲   badge: … ✓ ✕ !                       │                                   macOS notification
        └───────────────────────────────────────┘                                     "✅ <filename>"

• daemon — Python standard-library http.server, listens on 127.0.0.1:7788, started at login by launchd.
• extension — Chrome MV3, a single toolbar button, grabs tab.url and POSTs it to the daemon.
• output — ~/Downloads/<platform>_<title>_<date>.mp4 (≤1080p mp4 by default; configurable).
• log — ~/Library/Logs/xiaoer-videolab.log

✅ What you can download

Powered by yt-dlp's 1872 extractors — most video sites work. A practical map:

	Sites
✅ Tested & confirmed	YouTube · Vimeo · Bilibili (B站) · Douyin (抖音) · Xiaohongshu (小红书)
✅ Supported (yt-dlp extractor, same path)	X/Twitter · Ixigua (西瓜) · Instagram · Reddit · Dailymotion · Facebook · TikTok* · …and ~1860 more
⚠️ Free content only	Youku (优酷) · iQiyi (爱奇艺) — VIP / DRM-protected episodes can't be downloaded
🚫 Not recommended	Weibo (微博) · Zhihu (知乎) — see note below
❌ Not supported	Kuaishou (快手) & Tencent Video (腾讯视频) — no extractor; WeChat Channels (视频号) — in-app & encrypted

🚫 Weibo / Zhihu — not recommended. Their web pages are combined SPA feeds (much like TikTok): the video is just one small part of a big page, and you usually can't open a single video on its own URL. With no clean per-video address to grab, the button has nothing reliable to work with — so we suggest skipping them.

* TikTok / overseas sites need a network that can reach them (a proxy in mainland China; note some datacenter IPs are blocked by TikTok's API).

🎯 视频号 / 快手 / 小程序 / 直播流? Those live inside apps and need packet-sniffing — use res-downloader for them. This tool focuses on the yt-dlp universe.

抖音 & 小红书 use a special in-page grabber (yt-dlp can't read them), so click the button while the video is open/playing on the page.

Notes: 平台 (platform) and 标题 (title) are auto-detected for the filename; 日期 (date) is the download day.

TL;DR (if you've done this before)

brew install yt-dlp ffmpeg
git clone https://github.com/Jane-xiaoer/xiaoer-videolab.git
cd xiaoer-videolab && ./scripts/install.sh
# then load extension/ as an unpacked extension at chrome://extensions/

---

Installation — step by step

First time? Follow every step below. It takes about 5 minutes and you only do it once.

What you need

• macOS (background service via launchd) or Windows 10/11 (via Task Scheduler).
• A Chromium-based browser — Chrome, Arc, Edge, Brave, or Dia.
• About 5 minutes.

On Windows? Parts A & B below are for macOS. Jump to 🪟 On Windows, then come back for Part C.

You do not need to know how to code. You will copy-paste a few commands.

Part A · Install the engine (one-time)

This tool is a friendly button in front of yt-dlp, the open-source downloader that does the real work. So you install that first.

A1. Open the Terminal app. (Press ⌘ Space, type Terminal, hit Enter.)

A2. Install Homebrew (a package manager for Mac). If you already have it, skip to A3. Paste this into Terminal and press Enter, then follow its prompts:

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

A3. Install yt-dlp and ffmpeg:

brew install yt-dlp ffmpeg

ffmpeg matters — without it some sites give you audio-only or low quality, because the video and audio come as separate streams that ffmpeg merges back together.

Part B · Install Xiaoer VideoLab

B1. Still in Terminal, paste these three lines:

git clone https://github.com/Jane-xiaoer/xiaoer-videolab.git
cd xiaoer-videolab
./scripts/install.sh

B2. When it works, the last lines you see will look like this:

✓ Daemon running at http://127.0.0.1:7788
  Log: ~/Library/Logs/xiaoer-videolab.log

Next: load the browser extension
  1. Open chrome://extensions/
  ...

That means the background downloader is installed and will start automatically every time you log in. You never touch the Terminal again.

🪟 On Windows? (do this instead of Parts A & B)

The browser steps (Part C & D) are identical on every OS — only the engine + service install differs.

W1. Install the engine. Open PowerShell and run (uses winget, built into Win 10/11):

winget install Python.Python.3.11 yt-dlp.yt-dlp ffmpeg

W2. Install Xiaoer VideoLab. Either grab the code with the green Code → Download ZIP button (then unzip), or in PowerShell:

git clone https://github.com/Jane-xiaoer/xiaoer-videolab.git
cd xiaoer-videolab
powershell -ExecutionPolicy Bypass -File scripts\install.ps1

The installer registers a login-start task and asks which browser to pull cookies from (edge/chrome). To uninstall later: powershell -ExecutionPolicy Bypass -File scripts\uninstall.ps1.

Now continue with Part C below — loading the toolbar button is the same everywhere.

Part C · Add the toolbar button

The browser button isn't on the Chrome Web Store (yet), so you load it manually. This is normal and safe.

The screenshot below marks every step in this part (labels are in Chinese, but the red arrows point at exactly the right buttons): ① open the address · ② turn on Developer mode · ③ click "Load unpacked" · ④ flip the extension's switch on.

C1. Open a new browser tab and go to: chrome://extensions/ (On Edge it's edge://extensions/, on Arc/Brave the same chrome://extensions/.)

C2. Turn on Developer mode — the switch in the top-right corner.

C3. Click the Load unpacked button (top-left). A folder picker opens.

C4. Navigate to where you cloned the repo and select the extension folder inside it (e.g. xiaoer-videolab/extension). Click Select.

A card titled Xiaoer VideoLab now appears in your extensions list.

C5. Click the puzzle-piece icon in your toolbar, find Xiaoer VideoLab, and click the pin so its icon stays on the toolbar.

Part D · Download your first video

D1. Open any video page (YouTube, Bilibili, X, TikTok …).

D2. Click the Xiaoer VideoLab icon in your toolbar.

D3. You'll get a "Downloading…" notification, then a "✅ <filename>" notification when it's done. The icon also shows a little badge:

Badge	Meaning
…	request sent, downloading
✓	daemon accepted the job (download continues in the background)
✕	can't reach the daemon
!	daemon returned an error (check the notification / log)

D4. Find your video in the ~/Downloads folder. 🎉

That's it — from now on it's just open a video → click the button.

Configuration

All optional — set them and re-run ./scripts/install.sh to bake them into the service:

Variable	Default	What it does
VIDEOLAB_PORT	7788	daemon port — ⚠️ if you change it, also edit extension/background.js (DAEMON) and extension/manifest.json (host_permissions) to the same port, or the button can't reach the daemon
VIDEOLAB_DOWNLOADS	~/Downloads	where files land
VIDEOLAB_YT_DLP	auto-detect	path to the yt-dlp binary (auto-detect prefers a yt-dlp-nightly build if one is installed — see FAQ on HTTP Error 412)
VIDEOLAB_PREFIX	(none)	filename prefix, e.g. 小耳-
VIDEOLAB_MAX_HEIGHT	1080	max video height (set 2160 for 4K)
VIDEOLAB_COOKIES_BROWSER	(off)	pull cookies from a browser (chrome/brave/firefox/edge/safari) for login-gated / private videos
VIDEOLAB_APP_NAME	Xiaoer VideoLab	name in notifications

# example: 4K, pull login cookies from Chrome, brand the filenames
VIDEOLAB_MAX_HEIGHT=2160 VIDEOLAB_COOKIES_BROWSER=chrome VIDEOLAB_PREFIX="小耳-" ./scripts/install.sh

Commands

# is the daemon alive?
curl http://127.0.0.1:7788/health

# tail the log
tail -f ~/Library/Logs/xiaoer-videolab.log

# restart the daemon
launchctl unload ~/Library/LaunchAgents/com.xiaoer.videolab.plist
launchctl load   ~/Library/LaunchAgents/com.xiaoer.videolab.plist

# download without the extension
curl -X POST http://127.0.0.1:7788/download \
  -H 'Content-Type: application/json' \
  -d '{"url":"https://www.youtube.com/watch?v=dQw4w9WgXcQ"}'

# update everything (latest code + newest yt-dlp engine), keeping your settings
./scripts/update.sh

# uninstall the service
./scripts/uninstall.sh

🔄 The yt-dlp engine auto-updates weekly (a launchd job, Sundays ~4am — log: ~/Library/Logs/xiaoer-videolab-ytdlp-update.log), so fast-moving sites keep working without you doing anything. Run bash scripts/auto-update-ytdlp.sh any time to update it now. (Code updates to VideoLab itself stay manual via ./scripts/update.sh.)

Security

• The daemon binds to 127.0.0.1 only — it is not reachable from your network.
• No drive-by downloads. /download rejects any request carrying an http(s) Origin header, so a malicious web page's JavaScript cannot make the daemon download files behind your back. The extension (chrome-extension://) and command-line calls (no Origin) are allowed.
• The extension's only host permission is http://127.0.0.1:7788/*. It reads only the current tab's URL when you click — no page content, no content scripts.
• If you also want to block other local processes, add a shared X-Token header check in daemon/server.py.

FAQ

It says "can't reach the daemon" (✕). Run curl http://127.0.0.1:7788/health. If that fails, check tail ~/Library/Logs/xiaoer-videolab.err.log and confirm yt-dlp is installed.

A video downloads at low quality / audio only. Some sites split streams; make sure ffmpeg is installed so yt-dlp can merge them.

A private / members-only video fails. Set VIDEOLAB_COOKIES_BROWSER to the browser where you're logged in, then re-install.

A site that used to work now fails — Bilibili returns HTTP Error 412, or a site throws extractor errors. The site tightened its anti-bot defenses and your yt-dlp is older than the fix. Update it first (yt-dlp --update, or brew upgrade yt-dlp). Stable releases can lag days-to-weeks behind fast-moving sites like Bilibili — if updating stable isn't enough, install the nightly build, which VideoLab auto-detects and prefers over stable:

# self-contained nightly binary (macOS) — VideoLab picks it up automatically, no config needed
mkdir -p ~/.local/bin
curl -L -o ~/.local/bin/yt-dlp-nightly \
  https://github.com/yt-dlp/yt-dlp-nightly-builds/releases/latest/download/yt-dlp_macos
chmod +x ~/.local/bin/yt-dlp-nightly
# update it any time it falls behind again:
~/.local/bin/yt-dlp-nightly --update-to nightly

Not on macOS? The extension is cross-platform; the installer is macOS-only. On Linux/Windows just run python3 daemon/server.py yourself (any process manager works).

Contributing

Issues and PRs welcome. The whole thing is ~400 lines of stdlib Python + vanilla JS — easy to read, easy to fork. If you add support for a workflow you care about (a new format profile, a Firefox manifest, a Linux service file), send it over.

Author

Jane · 小耳 / Xiaoer — a family of little tools that listen, read, find, and organize.

• GitHub: @Jane-xiaoer
• Email: xiaoerzhan@gmail.com

Part of the Xiaoer toolbox, alongside Xiaoer Ask and Smart Rename.

📱 关注作者 / Follow Me

如果这个仓库对你有帮助，欢迎关注我。后面我会持续更新更多 AI Skill、做网站、自动化工作流和创意项目。

If this repo helped you, follow me for more AI skills, website building, automation workflows, and creative projects.

• X (Twitter): @xiaoerzhan
• 微信公众号 / WeChat Official Account: 扫码关注 / Scan to follow

中文：欢迎关注我的公众号，一起研究 AI Skill、网站搭建、自动化流程和创意实验。

English: Follow my WeChat Official Account for more AI skills, website building, automation workflows, and creative experiments.

Acknowledgements

Standing entirely on the shoulders of yt-dlp — this project is just a friendly one-click button in front of it. Please support and respect the yt-dlp project.

License

MIT © 2026 Jane (小耳 / Xiaoer)

Download only content you have the right to download. You are responsible for respecting the terms of service of the sites you use this on, and applicable copyright law.