Extract the audio from a video with ffmpeg from command line A developer built a reusable ffmpeg script, packaged in the cmdbook tool, that extracts audio from video files by first probing the embedded codec with ffprobe and selecting a matching output extension, avoiding the common failure of copying AAC audio into an .mp3 container. The script offers an 'original' mode that copies the audio stream without re-encoding, extracting a 20-minute video's audio in under a second, while mp3 re-encoding took about 3 seconds. Marking the input file and format as placeholders gives the script a form-based GUI, with overwrite prompts and a no-audio-track check. You have a recording of a call or a talk, and you only want the audio. How do you do it? You ask an AI, and it gives you the classic line: ffmpeg -i "Team call.mp4" -vn -acodec copy out.mp3 It fails: Invalid audio stream. Exactly one MP3 audio stream is required. Could not write header incorrect codec parameters ? : Invalid argument and leaves an empty out.mp3 behind. The audio inside an .mp4 is usually AAC, and copy can't put AAC into an .mp3 . You ask again. Now you either re-encode to mp3, or copy into .m4a , which only works because this video happens to have AAC inside. So you save the script somewhere: a Markdown note, or an extract-audio.sh file. Next time it's a .webm with different audio inside, and the extension in your script is wrong again. With cmdbook https://printf.app/cmdbook , the script has one place to live, and it gets a GUI for free. 1. Save the script once. Create a command in cmdbook and paste the script it's at the end . It asks ffprobe what audio is inside, then picks the right extension. 2. Mark what changes. The video and the format become placeholders: IN="{{prompt:video}}" FORMAT={{prompt:format?original|mp3|m4a|wav}} original keeps the audio as it is: no quality loss, and almost no time. 3. Click Run. cmdbook shows a form. Drag the video onto it, pick a format, press Enter. 4. Get the file. The audio is saved next to the video: php Audio track: aac - original 47M video - 19M Long meeting.m4a For a 20-minute video, original took under a second on my Mac, and mp3 took 3 seconds. If the file already exists, it asks before overwriting. If the video has no audio, it says so. Next time, your last format is already selected. That's all a script needs to get a GUI: mark the parts that change. The rest is the script you already have. Try cmdbook here: printf.app/cmdbook https://printf.app/cmdbook The full script macOS, needs ffmpeg IN="{{prompt:video}}" FORMAT={{prompt:format?original|mp3|m4a|wav}} export PATH="/opt/homebrew/bin:/usr/local/bin:$PATH" command -v ffmpeg >/dev/null || { echo "ffmpeg not found. Install it: brew install ffmpeg"; exit 1; } -f "$IN" || { echo "Not a file: $IN"; exit 1; } What audio is inside? "original" keeps it as is fast, no quality loss , so the file extension has to match the codec. CODEC=$ ffprobe -v error -select streams a:0 -show entries stream=codec name -of csv=p=0 "$IN" if -z "$CODEC" ; then echo "No audio track in $IN." exit 1 fi case "$FORMAT" in original case "$CODEC" in aac|alac EXT=m4a ;; mp3 EXT=mp3 ;; opus EXT=opus ;; vorbis EXT=ogg ;; flac EXT=flac ;; pcm EXT=wav ;; EXT=mka ;; esac CODEC ARGS="-c:a copy" ;; mp3 EXT=mp3; CODEC ARGS="-c:a libmp3lame -q:a 2" ;; m4a EXT=m4a; CODEC ARGS="-c:a aac -b:a 192k" ;; wav EXT=wav; CODEC ARGS="-c:a pcm s16le" ;; echo "Unknown format: $FORMAT"; exit 1 ;; esac OUT="${IN%. }.$EXT" if -e "$OUT" ; then printf "%s already exists. Overwrite? y/N " "$OUT" read -r ANSWER case "$ANSWER" in y|Y|yes ;; echo "Left as is."; exit 0 ;; esac fi echo "Audio track: $CODEC -> $FORMAT" if ffmpeg -nostdin -hide banner -v error -stats -y -i "$IN" -map 0:a:0 $CODEC ARGS "$OUT"; then rm -f "$OUT" exit 1 fi echo "$ du -h "$IN" | cut -f1 video -> $ du -h "$OUT" | cut -f1 $OUT"