antonio-orionus/Arroxy

Media & Design

A desktop app for downloading video and audio from YouTube and 2000+ sites, with a built-in browser for signing in to private or restricted content.

779
99
TypeScript
GPL-3.0
6 months old

Worth a try if you download from YouTube often and want a desktop GUI that handles sign-in and playlists without exporting cookies. The trade-off is an Electron app with a manual first-launch step on Windows.

The problem

Someone wants an archived copy of a video, a playlist, or just the audio track at good quality. The common yt-dlp path means installing a CLI, remembering flags, and for account-only content, exporting cookies from a browser and feeding them to the tool. Arroxy moves that into a desktop window: you paste a link or browse inside the app, pick a format, and the file lands in a folder you chose. The account-bound material is where the workflow actually changes, because you sign in on the site's own page inside the app instead of exporting cookies.

What's different

The central mechanism is Explore, a browser embedded in the app. You can right-click a link to download it, or use the Download button for the page you are on, and a sign-in made there is used only for the matching site. The README says the app reads the sign-in fresh for a download and deletes its copy afterwards, and that signing out in Settings forgets it. That covers private videos, Watch Later and Liked lists, and age-restricted material without a separate cookie file. The same reason makes it heavier than a CLI wrapper: it is an Electron application, so it carries the footprint of a full browser.

Underneath, downloads go through yt-dlp, so format coverage follows what the source site serves: the README lists up to 8K, 60 fps and HDR streams, AV1/VP9/H.264 choices at each resolution, and audio-only MP3, M4A/AAC, Opus or WAV. Extras include subtitles, SponsorBlock, partial-video downloads and 50 interface languages. Whether those extras work well across the claimed 2000+ sites is not something the README shows; the examples lean on YouTube.

Who should try it

This fits someone who downloads regularly on a desktop, cares about quality, and sometimes needs content tied to their own account. It is also a reasonable fit for people who would rather browse and click than construct command lines. It is a poor fit for headless servers, batch pipelines or remote automation, since the value lives in the GUI and the embedded browser. Linux users have AppImage, Flatpak and tar.gz options, but builds are x86_64 only, as the install page states.

Requirements and limitations

Windows 10/11, macOS 13+ (Intel or Apple Silicon), or Linux, all x86_64. The Windows installer is unsigned, and the README warns that Defender heuristics sometimes quarantine the NSIS setup or the portable build, with steps to restore it and add an exclusion; treat that as a first-launch hurdle, not a background risk to ignore. If the window does not appear, the README's troubleshooting section points to a GPU hang and offers a --disable-gpu launch flag plus a persistent argv.json setting.

Ongoing care matters more than installation. yt-dlp and site extraction logic need updating as sites change, so a downloader that works today can break after a site change until the app ships an update. The bundled browser also means account sessions live inside the app; the README recommends a secondary account for Explore and describes deleting sign-in copies after use, but users should still sign out deliberately. GPLv3 licensing matters if you intend to redistribute or fork.

HotRepo's take

The only dated HotRepo observation is a weekly TypeScript board entry on 2026-10-24: +172 stars that week, rank #13. There is no daily-board history, so persistence cannot be judged from that single sample. At 183 days old, the project is young; the weekly rank shows current attention, not a track record. Before committing, check the release notes for yt-dlp update cadence, since that is what determines whether downloads keep working after site changes.

Written by AI from the project's README, updated October 25, 2026. The project's own docs are the source of truth.