ffmpeg symlink to rffmpeg fails version test when starting Jellyfin

[gotkube]

Running Jellyfin 10.10.6.0 on a Slackware system.
Followed the instructions in the SETUP.md file; adjusting for Slackware instead of Debian.
Deviated a bit by using a venv for rffmpeg and a wrapper script (also called rffmpeg) to run rffmpeg via the venv; symlinked ffmpeg and ffprobe to that wrapper script instead of the downloaded Python rffmpeg script. (wanted to use a venv so I can build it into a Slackware package to put on other machines)

I can run rffmpeg via the wrapper script, and used it to call the 'init' command and add hosts. Seems like rffmpeg is working.

When I start Jellyfin, it does an ffmpeg validation check that fails:

[19:26:19] [ERR] [1] MediaBrowser.MediaEncoding.Encoder.MediaEncoder: FFmpeg validation: The process returned no result
[19:26:19] [ERR] [1] MediaBrowser.MediaEncoding.Encoder.MediaEncoder: FFmpeg: Failed version check: /usr/local/bin/ffmpeg
[19:26:19] [ERR] [1] MediaBrowser.MediaEncoding.Encoder.MediaEncoder: FFmpeg: Path set by command line or environment variable is invalid
[19:26:19] [FTL] [1] Main: Error while starting server
MediaBrowser.Common.FfmpegException: Failed to find valid ffmpeg

Using the tests mentioned in SETUP.md should be returning an ffmpeg version as if it was ffmpeg itself, but instead it returns:

> /usr/local/bin/ffmpeg -version
Usage: rffmpeg [OPTIONS] COMMAND [ARGS]...
Try 'rffmpeg -h' for help.

Error: No such option: -v

I don't understand why this is failing the version test. Can I not use a venv or wrapper script for rffmpeg?

[Freak4915]

I am also experiencing the same, however, my setup is an Ubuntu 24.04 VM, followed the setup instructions with the only difference being the user. Can run the symbolic link ffmpeg and ffprobe and it returns a valid version, however, jellyfin still fails to start stating "Failed to find valid ffmepg". Reverting back to jellyfins native ffmpeg and ffprobe, the service starts.

The versions between the native and rffmpeg match "ffmpeg version 7.1.1-Jellyfin Copyright (c) 2000-2025 the FFmpeg developers"

Jellyfin must be running a new check which is failing if run via rffmpeg, but I haven't been able to locate.

May attempt installing an earlier version of jellyfin and see what might have changed.

Jellyfin.Server 10.10.7.0 currently

[joshuaboniface]

@gotkube I suspect there's something going on with your wrapper; I can't replicate this with rffmpeg as-is, and it looks like it's not properly detecting which subcommand to use. rffmpeg will determine what command it should be emulating based on the $0 value, which is probably being lost in the wrapper somehow.

@Freak4915 This is likely unrelated unless you are also using a wrapper. I would need to see debug logs from both Jellyfin and rfmpeg to know why it might be failing.

[emiliopedrollo]

A very similar thing is happening to me. Same logs on Jellyfin, however I can run ffmpeg -verion and it returns the version from the transcode node:

2026-01-01 08:20:22,275 - rffmpeg[2405] - INFO - Starting rffmpeg as /usr/local/bin/ffmpeg with args: -version
2026-01-01 08:20:22,275 - rffmpeg[2405] - DEBUG - Determining optimal target host
2026-01-01 08:20:22,275 - rffmpeg[2405] - DEBUG - Using SQLite as database.
2026-01-01 08:20:22,301 - rffmpeg[2405] - DEBUG - SQLite connection closed.
2026-01-01 08:20:22,301 - rffmpeg[2405] - DEBUG - Using SQLite as database.
2026-01-01 08:20:22,301 - rffmpeg[2405] - DEBUG - SQLite connection closed.
2026-01-01 08:20:22,301 - rffmpeg[2405] - DEBUG - Trying host ID 1 '192.168.1.100'
2026-01-01 08:20:22,301 - rffmpeg[2405] - DEBUG - Running SSH test
2026-01-01 08:20:22,708 - rffmpeg[2405] - DEBUG - SSH test succeeded with retcode 0
2026-01-01 08:20:22,708 - rffmpeg[2405] - DEBUG - Selecting host as idle
2026-01-01 08:20:22,708 - rffmpeg[2405] - DEBUG - Found optimal host ID 1 '192.168.1.100' (node1)
2026-01-01 08:20:22,708 - rffmpeg[2405] - INFO - Running command on host '192.168.1.100' (node1)
2026-01-01 08:20:22,708 - rffmpeg[2405] - DEBUG - Remote command: /usr/bin/ssh -q -t -o ConnectTimeout=1 -o ConnectionAttempts=1 -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null -o ControlMaster=auto -o ControlPath=/run/shm/ssh-%r@%h:%p -o ControlPersist=300 -i /config/.ssh/id_ed25519 -p 2022 jellyfin@192.168.1.100 /usr/lib/jellyfin-ffmpeg/ffmpeg -version
2026-01-01 08:20:22,708 - rffmpeg[2405] - DEBUG - Using SQLite as database.
2026-01-01 08:20:31,998 - rffmpeg[2405] - DEBUG - SQLite connection closed.
2026-01-01 08:20:32,015 - rffmpeg[2405] - DEBUG - Cleaning up after signal  PID 2405
2026-01-01 08:20:32,016 - rffmpeg[2405] - DEBUG - Using SQLite as database.
2026-01-01 08:20:37,730 - rffmpeg[2405] - DEBUG - SQLite connection closed.
2026-01-01 08:20:37,730 - rffmpeg[2405] - INFO - Finished rffmpeg with return code 0

However, when Jellyfin tries to validate it there's absolutelly no log.

[ChunkyPanda03]

I was able to get it working in my docker container I needed to chmod/chown the rffmpeg.log and the .db file I can link my docker container if anyone is interested I would just need to push it to github

[irrelevantdotcom]

I have the exact same issues... Modified the Dockerfile to add symlinks from /var/lib/jellyfin-ffmpeg/ff* to /usr/local/bin/rffmpeg, but now get the ffmpeg fails version check on startup.

I note from the (jellyfin) source there's a configuration key
public const string FfmpegSkipValidationKey = "FFmpeg:novalidation";
which disables the test, but I have no idea where to set this..

[irrelevantdotcom]

Answering my own post. Set an environment variable! In docker-compose.yml it's:
- JELLYFIN_FFMPEG__NOVALIDATION=true
Note the double underscore.

[PassiveLemon]

I was also getting this too, but I believe I've found the cause.

I can successfully ssh into my transcode server:

[lemon@titanium:~]$ docker exec -it jellyfin ssh -i /home/docker/.ssh/id_ed25519 docker@silver
[docker@silver:~]$ ffmpeg
ffmpeg version 7.1.2-Jellyfin Copyright (c) 2000-2025 the FFmpeg developers
built with gcc 15.2.0 (GCC)

but using docker exec -it jellyfin /usr/lib/jellyfin-ffmpeg/ffmpeg -version will give the reported error.

I did some digging and the logic depends on how you call it:

if __name__ == "__main__":
    all_args = sys.argv
    cmd_name = all_args[0]

    # Load the config
    config = load_config()

    if "rffmpeg" in cmd_name:
        run_control(config)

    else:
        ffmpeg_args = all_args[1:]
        run_ffmpeg(config, ffmpeg_args)

It checks if "rffmpeg" is in sys.argv[0], which is the path/name of the process so if your symlink to rffmpeg is something like ffmpeg -> /usr/bin/rffmpeg, then the if "rffmpeg" in cmd_name: check will pass because of "rffmpeg" being in the path name, which means it always runs in control and doesn't correctly pass the ffmpeg calls.

What I did was just rename rffmpeg to r-ffmpeg everywhere and now I no longer get that issue:

[lemon@titanium:~]$ docker exec -it jellyfin /usr/lib/jellyfin-ffmpeg/ffmpeg -version
ffmpeg version 7.1.2-Jellyfin Copyright (c) 2000-2025 the FFmpeg developers
built with gcc 15.2.0 (GCC)

Final compose:

jellyfin:
  image: lscr.io/linuxserver/jellyfin:latest
  container_name: jellyfin
  restart: on-failure:10
  volumes:
    - /data/docker/volumes/streaming/Jellyfin/:/config/:rw
    - /data/Media/Movies/:/data/movies/:ro
    - /data/Media/Shows/:/data/shows/:ro
    - /data/Media/Recordings/:/data/recordings/:ro
    # rffmpeg
    - /nix/store/:/nix/store/:ro # For /run/current-system/sw/bin/r-ffmpeg and ssh
    - /run/current-system/sw/bin/r-ffmpeg:/usr/lib/jellyfin-ffmpeg/ffmpeg:ro
    - /run/current-system/sw/bin/r-ffmpeg:/usr/lib/jellyfin-ffmpeg/ffprobe:ro
    - /data/docker/volumes/streaming/Jellyfin-rffmpeg/:/var/lib/rffmpeg/:rw
    - ./rffmpeg/rffmpeg.yml:/etc/rffmpeg/rffmpeg.yml:ro
    - /run/current-system/sw/bin/ssh:/usr/bin/ssh:ro
    - ./rffmpeg/keys/:/home/docker/.ssh/:rw
  environment:
    TZ: "${TZ}"
    PUID: "${PUID}"
    PGID: "${PGID}"
    JELLYFIN_FFMPEG__NOVALIDATION: "true"

@joshuaboniface I think this could be mostly fixed by just pulling out the basename from sys.argv[0]:

# Entrypoint
if __name__ == "__main__":
    all_args = sys.argv
    cmd_name = os.path.basename(all_args[0])

    # Load the config
    config = load_config()

    if "rffmpeg" in cmd_name:
        run_control(config)

    else:
        ffmpeg_args = all_args[1:]
        run_ffmpeg(config, ffmpeg_args)

That way it only runs when explicitly called as "rffmpeg"