Jellyfin vs Emby

In 2018, Emby relicensed at version 3.5.3 and moved its open components into plugins. A group of contributors forked 3.5.2, the last version under the old terms, and called it Jellyfin. Everything else on this page is a consequence of that afternoon.

Eight years later the family resemblance is thinner than people expect. Jellyfin rewrote its library layer, changed its plugin system and built its own client apps. Emby kept shipping on a closed codebase with a paid unlock. They still both default to ports 8096 and 8920, still both store a SQLite library database in a config directory, and still both feel like relatives when you use them. But the plugins do not cross over, the databases do not cross over, and the scripts do not cross over.

The real decision here is narrow: is there a specific problem you have that $119 makes disappear?

What the spec table is telling you#

Licence is the whole comparison. Everything else on this table is close enough to be noise: one container each, SQLite each, arm64 each, directory-copy backups each, and idle memory 280 MB against 350 MB, which nobody will notice.

Identity is the one other row worth stopping on. Emby is ldap, meaning a working LDAP plugin and no first-party OIDC, so anything provider-shaped means a forward-auth proxy that native clients then bypass. Jellyfin is oidc-plugin, which sounds better but is roughly as limited: the SSO plugin covers the web UI only, and its original repository (9p4) was archived in May 2026 with a community fork carrying it forward. If your household runs on a provider, LLDAP plus Emby's LDAP plugin is honestly the smoother of the two paths, and Single sign-on for self-hosters explains why neither is good.

What Emby Premiere actually costs and unlocks#

Premiere is roughly $4.99/month, around $54/year, or $119 lifetime. It unlocks hardware transcoding, mobile sync (downloads for offline viewing), cover art, DVR and the full mobile apps.

Every one of those is free in Jellyfin. That is not a rhetorical point, it is the entire pitch: Jellyfin has no paid tier, no account and no licence check anywhere.

Where Emby's model is defensible is against Plex Media Server, not against Jellyfin. Plex now charges a platform fee, and since 29 April 2025 remote playback of your own files requires someone in the chain to hold a pass, enforced across Fire TV, Samsung, LG, PlayStation and Xbox from 29 April 2026. Emby asks for a licence, not permission: nothing about your server requests approval from a vendor cloud to play a file on your own network. If you are leaving Plex over the terms rather than the price, Emby is a coherent destination and Jellyfin vs Plex covers that fork in the road in detail.

If you are paying anyway, the arithmetic is short: $119 Emby lifetime, $249.99 Plex lifetime, $0 Jellyfin. Compare all three before you assume the middle option is a compromise.

Which one your household will complain about less#

Honestly, it depends on the screen, and Emby's advantage is smaller than its price implies.

Emby's apps are more consistent than Jellyfin's on some TV platforms and its metadata handling has fewer sharp edges out of the box. Jellyfin's web client, Android TV app and Kodi integration are strong; Roku and older Samsung and LG sets are where it still shows its seams, along with Dolby Vision profile 7 and lossless TrueHD passthrough on specific televisions.

The uncomfortable part for Emby is the ecosystem. It has fewer third-party clients than Plex and fewer than Jellyfin on some TV platforms, so if the app on your particular TV is the weak one, there is no community alternative to fall back to and no source to fix it with. Jellyfin's client gaps get filled eventually by somebody. Emby's do not get filled by anybody but Emby.

Request tooling is a wash: Jellyseerr supports Jellyfin, Emby and Plex.

Which one gets hardware transcoding working on your hardware#

Both use the same underlying paths. The setup differs in two ways that cost people evenings.

Jellyfin needs the device node and the render group:

services:
  jellyfin:
    image: jellyfin/jellyfin:10.11.11
    devices: [ /dev/dri:/dev/dri ]
    group_add: [ "989" ]   # getent group render | cut -d: -f3

Emby needs the device node and GIDLIST, because its entrypoint sets up its own user and ignores compose's group_add:

services:
  emby:
    image: emby/embyserver:latest      # emby/embyserver_arm64v8 on ARM
    devices: [ /dev/dri:/dev/dri ]
    environment:
      - GIDLIST=44,989                 # video,render on the host

Copy a Jellyfin fragment into an Emby stack and device access fails quietly. And on Emby none of it does anything without an active Premiere key: the acceleration options appear in Settings, transcodes still run on CPU, and you can lose an evening to VAAPI before checking the licence page. On ARM64 that bites hardest, because software transcoding on an SBC is usually not viable at all. GPUs, transcoding and local AI covers what each GPU actually delivers in simultaneous streams.

What happens on upgrade day#

Emby wins this one. The 4.8 and 4.9 series have been undramatic tag pulls; current stable is 4.9.5.0 (18 May 2026) with 4.10 in beta.

Jellyfin's 10.11 (October 2025) was the hardest upgrade in its history. It moved the library onto EF Core and merged library.db into a single jellyfin.db. The migration is supported only from 10.10.7, runs once, has no downgrade, and on libraries past roughly 50,000 items can churn for hours with no visible progress. People who killed the container mid-migration lost watch history. It also dropped armhf outright, stranding 32-bit Raspberry Pi installs on 10.10.x. Current stable is 10.11.11 (June 2026); the next major skips 11 and goes to 12.0, in release candidates since June 2026.

Backups are near-identical work: copy the whole config directory (/config, or /var/lib/emby on a package install), stop the container first or use Jellyfin's scheduled backup task from 10.11, and exclude the transcode temp and cache. One Emby detail to write down: the Premiere key is tied to your account, not the config directory, so a restore onto new hardware needs the key re-entered. See Backups that actually restore and An update strategy that does not lose data.

Which one you can fix when it breaks#

This is the argument that should carry more weight than it usually does.

Emby's builds are not reproducible and the build scripts are closed. If the project changes direction, raises the price, changes what Premiere covers, or abandons the platform your TV runs, there is no fork, no patch and no audit. That is exactly the situation that produced Jellyfin in 2018, and the people who lived through it are the reason Jellyfin exists.

Jellyfin's risks are different and worth naming rather than glossing: volunteer client maintainers stop, plugins get archived, and majors can migrate your database one way. Those are project risks you can work around, fund or fix. They are not someone else's licensing decision applied to your library.

The verdict, by situation#

What to do next#

Install Jellyfin first, because it is free and takes twenty minutes. Point it at your media read-only, get hardware transcoding confirmed in the playback info overlay, and watch on the actual televisions in your house for a week. If nothing annoys you, you are done and you saved $119.

If something does annoy you, write down what it is before you install Emby. "The Roku app is slow" is a reason to try Emby. "It felt less polished" usually means you will pay and stay slightly annoyed. Emby offers a trial, and both servers happily read the same media directory at the same time on different ports, so run them side by side rather than migrating twice.

While you are there, sort out remote access properly with Remote access without port forwarding, and use Stack planner to check the whole media stack fits in your RAM. The rest of the category is in Media.