The Audio tab gives you per-track gain and mute across dialogue, music, SFX and ambience.
Automatic ducking is derived from where dialogue actually sits on the timeline — not from a fixed interval. The music dips 0.3 seconds before each line lands and releases 0.5 seconds after, at −9dB. Ducking that reacts late is audibly wrong. Events closer than 1.2 seconds apart are merged so the bed does not pump.
Loudness targets: streaming −24 LUFS, web −14, social −14, cinema −27. Pick one and the mix plan is recalculated.
An honest limit: without FFmpeg the app cannot decode audio, so it does not report a measured LUFS. It estimates from mix structure — track count, gains, dialogue density — and marks the result measured: false with the reason. A fabricated LUFS number is the kind of false precision that gets a master rejected on delivery.
You can view the exact FFmpeg filtergraph the mix would apply, and run it yourself.