G: Here’s a safe, no-reencode extractor. It pulls the first audio track from every .mkv and saves it as a Matroska audio file (.mka), which Jellyfin can use as an external audio track.

Simple (same folder, same basename)

shopt -s nullglob
for f in *.mkv; do
  base="${f%.*}"
  out="${base}.external.mka"
  echo ">> $f -> $out"
  ffmpeg -y -i "$f" -map 0:a:0 -c copy "$out"
done

With language code in the filename (helps Jellyfin label it)

shopt -s nullglob
for f in *.mkv; do
  base="${f%.*}"
  # Get language tag of first audio stream (falls back to 'und')
  lang=$(ffprobe -v error -select_streams a:0 \
         -show_entries stream_tags=language \
         -of default=nw=1:nk=1 "$f")
  [ -z "$lang" ] && lang="und"
  out="${base}.${lang}.external.mka"
  echo ">> $f -> $out"
  ffmpeg -y -i "$f" -map 0:a:0 -c copy "$out"
done

Notes

.mka keeps the original audio codec without re-encoding, so it’s fast and lossless for this step.
Put the resulting .mka files next to their videos. Using the same base name (optionally with a language code like .eng) helps Jellyfin auto-associate them.
If a file has no audio stream tagged as a:0, ffmpeg will error. You can skip such files by checking with ffprobe first if needed.

2025-09-10 11:56:35 EDT

M: what about if, in another folder, i wanted a loop that would look at every mkv file and, if there's a similarly named mka file, would alter the mkv so that the mka became the first audio track, with the original shifted to the second audio track?

G: [Non-text content: thoughts; content_type=thoughts]

G: [Non-text content: reasoning_recap; content_type=reasoning_recap]

G: Got it. Here’s a robust ffmpeg-based loop that, for every .mkv in the current folder, looks for a same-basename .mka (optionally in a separate MKA_DIR) and remuxes so that the .mka becomes audio #0 (default) and the original audio track(s) are shifted after it. All other streams (video, subs, chapters) are preserved losslessly.

Usage

Put this script in the folder with your .mkv files.
Set MKA_DIR to the folder containing your .mka files (or leave empty to use the same folder).
It writes to a temp file and atomically replaces the original .mkv after success.

#!/usr/bin/env bash
set -euo pipefail
shopt -s nullglob

# Set this if your .mka files live elsewhere, e.g. MKA_DIR="/path/to/mka"
MKA_DIR="${MKA_DIR:-}"

for v in *.mkv; do
  base="${v%.*}"
  a="${MKA_DIR:+$MKA_DIR/}$base.mka"

  if [[ -f "$a" ]]; then
    tmp="${base}.tmp.remux.mkv"
    echo ">> Injecting external audio for: $v"
    echo "   Using MKA: $a"

    # Inputs: 0 = original MKV, 1 = external MKA
    # Order streams so:
    #   - video/subs/data from 0 first (no audio yet)
    #   - audio from 1 (becomes a:0)
    #   - then original audio(s) from 0 (become a:1, a:2, ...)
    # Also copy chapters and clear any existing 'default' on audio before
    # marking the new a:0 as default.
    ffmpeg -hide_banner -loglevel error -y \
      -i "$v" -i "$a" \
      -map 0:v -map 0:s? -map 0:d? \
      -map 1:a \
      -map 0:a? \
      -map_chapters 0 \
      -c copy \
      -disposition:a 0 \
      -disposition:a:0 default \
      "$tmp"

    # Replace original atomically
    mv -f -- "$tmp" "$v"
    echo "   Done -> $v (external audio is now first/default)"
  else
    echo "-- Skipping $v (no matching $base.mka found)"
  fi
done

Notes

Completely lossless (-c copy), just a stream re-order + flag update.
Keeps all subtitles, attachments, data, and chapters (-map_chapters 0).
If a source has multiple original audio tracks, they’ll follow the new .mka in their original order.

If your external file sometimes has a different suffix (e.g., .eng.mka), you can change the line that sets a=... to a small glob like:

a_candidate=( ${MKA_DIR:+$MKA_DIR/}"$base".*.mka ${MKA_DIR:+$MKA_DIR/}"$base".mka )
a="${a_candidate[0]:-}"
[[ -n "${a:-}" && -f "$a" ]] || { echo "-- Skipping $v (no matching MKA)"; continue; }

This should slot right into your Jellyfin workflow.

2025-09-10 11:59:12 EDT

M: can you make that a cli command? the script looks good but i'd like to just paste it into the terminal

G: Absolutely 👍 — here’s a single-line command you can paste straight into your shell. It loops through .mkv files in the current folder, checks for a same-basename .mka, and remuxes so the .mka becomes audio #0 (default) and the original audio tracks are shifted after it:

for v in *.mkv; do base="${v%.*}"; a="${base}.mka"; if [[ -f "$a" ]]; then tmp="${base}.tmp.remux.mkv"; echo ">> Processing $v with $a"; ffmpeg -hide_banner -loglevel error -y -i "$v" -i "$a" -map 0:v -map 0:s? -map 0:d? -map 1:a -map 0:a? -map_chapters 0 -c copy -disposition:a 0 -disposition:a:0 default "$tmp" && mv -f -- "$tmp" "$v"; fi; done

That’s the whole script condensed to a one-liner for direct CLI use 🚀