Karaoke is a music system that plays a recorded accompaniment to a popular song. A performer sings the song live, usually by following the words on a video screen. The word also means the performance itself.
The Japanese word joins kara, which means empty, with a short form of the word for orchestra. So karaoke literally means empty orchestra.
In a metadata report, Karaoke is a field with a narrower meaning. The name comes from ffprobe, the file reader of the FFmpeg project.
FFmpeg lists each part of a file as a separate stream. Every stream gets a set of flags called the disposition, such as Karaoke and Attached Pic. The FFmpeg documentation defines this flag in a few words: the stream contains karaoke audio.
What the values look like
The -show_streams option of the ffprobe reader prints the flag for every stream, even in files with no music. The flag has 2 values in that output:
0: the flag is off1: the flag is on, so the stream is marked as karaoke audio
The exact form depends on the report. A JSON report shows "karaoke": 0, and a plain text report shows DISPOSITION:karaoke=0.
Which files carry it
FFmpeg reads and writes this flag in some formats but not in others, as these examples show:
- NUT: a stream can store the word
karaokein a tag calledDisposition - MKV: the Matroska specification has flags such as default, forced, and commentary, but no karaoke flag
- MP4: the FFmpeg list of MP4 track roles has no karaoke entry
- AC-3 and E-AC-3 audio: the audio data has a service type, and one value means karaoke
FFmpeg turns the karaoke service type into the flag when the audio has 2 or more channels, also in an MKV or MP4 file. With other audio formats, a stream in an MKV or MP4 file normally reports 0, even when the song is a karaoke version.
Some karaoke songs use a file type of their own. The Library of Congress lists the file extension kar for karaoke MIDI files.
Where the words of the song are stored
The Karaoke flag holds no words, only the number 0 or 1. FFmpeg has a separate Lyrics flag for a stream with song lyrics. The text itself sits in a different place in each format:
- MP3: the ID3v2 frame USLT holds the lyrics without timing, and the frame SYLT stores a time stamp with each syllable or text unit
- MKV: a
LYRICStag holds the words when the timing is not known, or as a copy of a subtitle track
Where the value comes from
FFmpeg sets the flags when it creates each stream, or later when it inspects the stream. For a NUT file, FFmpeg reads the flags from the Disposition tag. For AC-3 audio, FFmpeg reads the service type from the audio data.
With the FFmpeg option -disposition, you can set the flag yourself. By default, FFmpeg copies the flags from the input stream to the output stream.
A flag that only the container holds reaches the new file only if the new format can store it. The FFmpeg writer for MKV has no code for this flag.
Why it matters
A file can hold several audio streams, and this flag can mark one of them as karaoke audio. In FFmpeg commands, the stream selector disp:karaoke picks the streams that have the flag.
The flag is only a label, so it does not prove what you hear. A value of 0 does not prove that the song has no karaoke version or no lyrics.