Unboxing video music that stays under your voice: duck_db settings

For an unboxing video with a voiceover, set soundtrack.duck_db to roughly 8 to 12 and gain_db to a few dB below zero, then listen. Those are starting points to test, not measured values. The Timeline docs define duck_db as 0 to 20, default 0 (a static bed), implemented as a sidechain duck against the audio spine, and they reject it with audio.mode: "silence" as duck_requires_audio_spine. See the Timeline 1.0 page for the full schema.

Why unboxing needs ducking

An unboxing has long stretches of rustling, tape and short comments. A static bed at one level either drowns the comments or disappears during the quiet parts. Ducking lowers the music only while the voice is present, so the bed comes back up between lines. Because the duck follows the audio spine, the spine has to carry the voice and the box sounds you want kept.

Settings table

The music bed

Generate a light, curious, mid-tempo instrumental on the Music Router for a fixed $0.125 and avoid busy lead melodies that compete with speech. Google lists Lyria 3.5 at $0.08 per song on its pricing page, read on 2026-10-05; Sume bills $0.125 per accepted generation. A 3-minute unboxing renders at $0.10 per output minute, so $0.30, and you can run the unbilled plan call first to catch a bad fade value before paying.