Auto duck is a popular feature in audio mixing software like Audacity; it allows a track (A) to play at a desired volume while there is silence on another track (B). Whenever track B is NOT silent, track A automatically reduces its volume, allowing track B to be heard clearly, and both tracks to continue uninterrupted. A automatically returns to its default volume when B is silent once again, and so on for the duration of the track(s).

Example:

Track A: background music
Track B: voiceover (VO) track - spoken phrases separated by variable-length periods of silence.
This use case is common in meditation / relaxation skills, and games.
When there is speech on track B, the music volume (track A) reduces automatically to allow the voiceover on B to be heard clearly. Between spoken phrases, B is silent and the music volume (track A) returns to its default level.

Adding support for this in APLA would enable a much richer audio experience in many types of skill, including games, quizzes, education and relaxing sounds. One approach would be to add an an AutoDuck filter, applicable to items supported by the Mixer component.

{
  "type": "Audio",
  "source": "${payload.backgroundMusicUrl}",
  "filter": [
    {
      "type": "AutoDuck",
      "amount": 0.1
    }
  ]
}

Without this feature, ducking cannot be achieved natively by developers. They can only approximate the effect using pre-recorded audio, which makes it hard (impossible?) to mix a music track selected at runtime based on skill state, with spoken content pulled from a database.