Kyoo

Last reviewed Sep 13, 2026 · 877 words

I ran Kyoo alongside Jellyfin for 3 months to see whether the "modern Jellyfin" pitch held up, and the answer is that it holds up for anime and falls short for everything else. Kyoo is a 2,521-star GPL-3.0 media browser with the nicest interface in the category, automatic metadata that gets anime right where Jellyfin makes you fight, and a compose stack of about 10 containers that will not run on 512 MB. Its target user is specific: someone with a large anime and TV library, comfortable with Docker, who is tired of Jellyfin's scraper and does not need a mature client ecosystem yet.

What Kyoo gets right that Jellyfin does not

Two things. First, the interface. Kyoo's web front end looks like a 2020s streaming service: clean posters, a proper home page, per-show pages with seasons and specials handled sensibly, and a mobile app that shares the same codebase. Jellyfin is functional; Kyoo is pleasant. Second, anime. Kyoo's scanner parses filenames with guessit, understands absolute episode numbering, and matches against TheMovieDB and TVDB with anime-aware ordering, so a folder of 300 episodes named by absolute number lands in the right seasons on the first scan. On Jellyfin the same folder usually means a plugin, a rename pass with a tool, and manual identification for a third of the shows.

It also does hardware transcoding through a dedicated Go transcoder service, with environment switches for VAAPI, QSV and NVIDIA, and it transcodes on demand with a segment cache rather than pre-encoding. On a box with an Intel iGPU that behaves at least as well as Jellyfin's transcoder in my testing.

What it costs to run

The compose file from the repository starts a front end, a back end, the transcoder, a scanner, a matcher, an autosync service for scrobbling, PostgreSQL, Meilisearch for search, RabbitMQ for the internal queue, and Traefik to route them, exposed on port 8901. The catalogue's 1,024 MB minimum is the floor; budget 2 GB to be comfortable and more during a first scan of a large library. The single .env file carries the library path, the database credentials, and API keys for the metadata providers. Set LIBRARY_ROOT to your media, KYOO_URL to the public address, and read the rest before starting, because changing the database password after first boot is a manual fix.

This is a heavy stack for a media browser. Jellyfin is one container. If your server is a Raspberry Pi or a 2 GB VPS, Kyoo is not for you regardless of its merits.

The client situation is the real limitation

A media server is only as good as the devices that can play from it. Jellyfin has native apps for Android TV, Apple TV, Roku, Fire TV, iOS, Android, Kodi and a dozen community clients. Kyoo has its web app and its mobile app. On a television that means a browser or casting, and there is no Kyoo channel to install on the TV. For a household that watches on the sofa, this is the reason to keep Jellyfin as the primary server, and it is not a gap that closes quickly, because clients are the slowest thing to build in this space. The Jellyfin versus Plex question has an answer today; Kyoo is not yet in that conversation for living-room use.

Multi-user accounts exist with per-user watch state and an OIDC login option, so a shared server works. Parental controls and library-level permissions are thinner than Jellyfin's.

Running both is the sane pattern

Nothing stops two servers reading the same media folders read-only, and this is how I would introduce Kyoo: keep Sonarr and Radarr writing into the library, keep Jellyfin serving the TV, and add Kyoo for the anime fans in the house who watch on laptops and phones. Watch state does not sync between the two, which is the main annoyance, and Kyoo's autosync service scrobbles to Simkl, which partially fills that gap for people who track there. If in 6 months Kyoo has the TV clients and everyone has drifted to it, retire Jellyfin then.

Maturity, honestly

Kyoo has one primary developer, a small community, and release cadence that slows when that developer is busy. Breaking changes between versions have happened, the documentation is a README and a wiki rather than a docs site, and support means GitHub issues and a Discord. It is well-built and the architecture is sound, but it is a project to run for enjoyment, not one to bet a family's only media server on. The complete Jellyfin setup remains the default recommendation for a first server.

What I'd do

Keep Jellyfin. Add Kyoo on the same host with 2 GB reserved, the transcoder pointed at the iGPU, and only the anime and TV libraries mounted. Use it from laptops and phones for a season, and let the anime matching win people over on its own. If Kyoo grows native TV clients, revisit; until then it is the best-looking second server you can run, and a preview of what Jellyfin's interface should feel like.