The cheapest dB you will ever buy
Gain staging for voice-over: set it once, correctly
Gain staging for VO reduces to four moves: zero every software input slider, set your interface's gain so performance-volume speech peaks around −12 to −6 dBFS, get your mouth close enough that the gain can stay modest, and then leave all of it alone. Done once, correctly, it removes an entire family of noise problems before they exist.
Why gain is a noise decision, not a loudness decision
The gain knob amplifies everything that reaches it — your voice, your room, your preamp's own electronics — by the same amount. Turn it up and the ratio between voice and noise doesn't improve by a single dB; you've just made everything bigger, including the part that fails ACX's floor spec. The only lever that improves the ratio at record time is geometry: move the mouth closer and the voice rises while the room stays put. That's why the professional sequence is distance first, gain second — close the distance until the mic hears mostly you, then add only the gain needed to land peaks in the healthy zone. Every dB of gain you avoid is a dB of hiss and room you never have to remove; on the fix-order ledger, it's the cheapest noise reduction available at any price, which is why it sits at step three of the method.
Headroom between the target band and the ceiling is safety for the take where the character suddenly shouts.
The four-move setup
-
Zero the software sliders
Operating systems ship input-volume sliders and “enhancement” processing that silently stack on top of your interface's gain — a classic source of mystery hiss and pumping. Set OS input volume to its neutral position, switch off every OS-level enhancement, and make the interface knob the single place gain happens. One knob, one owner, no stacking.
-
Find performance volume, then set the knob
Read a real script at real energy — not a shy test sentence — and set interface gain so your meter peaks land around −12 to −6 dBFS, leaving headroom below the −3 dB ceiling ACX enforces and safety for the take where the character suddenly shouts. If your read has both whispers and shouts, stage for the shouts; whispers survive a boost far better than clipped peaks survive anything, because clipping is the one recording error with no undo.
-
Set distance before touching gain again
A hand-span from the mic, slightly off-axis, is the classic VO starting point: close enough that your voice dominates the room, far enough to manage plosives and proximity boom. If you need lots of knob to reach healthy peaks from there, that's information — a quiet mic, a gain-hungry dynamic, or a preamp working at the top of its range, which is where preamp hiss lives. The gain-drop diagnostic tells you whether your hiss is a staging artifact in one minute.
-
Freeze it and mark it
Tape at the knob position, a photo, a written settings card — anything that makes today's staging reproducible in six weeks, because chapter 14 must match chapter 2. ACX's requirement of consistent levels and tone across a whole book is, mechanically, a demand that your gain staging be boring and repeatable. Consistency also feeds measurement: your noise-floor number is only comparable over time if the gain behind it never wanders.
“Record low, boost in post” — the myth with a bill attached
The tempting version: record timid, stay miles from clipping, push levels up later in the DAW. The bill: a post boost raises voice and floor together, so a −70 dB floor under quiet takes becomes −58 after the +12 dB rescue — a floor that was passing now fails, and nothing about the room changed. Digital gain later is not the same as honest level now, because the noise you recorded is in the file forever, and every subsequent stage (reduction included) works harder on material that arrived faint. Stage for real levels at the source; post-gain is for trims, not rescues.
The boost lifts voice and floor together; the room never changed.
Levels for the live half
Calls have gain stages too — the app's input processing, automatic level control, sometimes a suppressor in the path — and the same one-owner principle applies. Let your interface set the level, keep app-side auto-adjust off where the app allows it, and let real-time suppression handle the room rather than cranking anything. A staged, suppressed call chain means the director hears a steady, present voice while your simultaneous local recording stays raw and correctly leveled — two outputs, both right, from one discipline.
Staging questions
- My interface has no meters — how do I hit −12 to −6?
- Use the recording software's meters; they read the same signal. Record a performance-volume passage, check where peaks landed, adjust the knob, repeat twice, and you're staged. Clip lights alone are not meters — by the time they speak, it's too late.
- Should compression happen while recording?
- For home VO, no — record clean and compress in post, where it's a choice instead of a commitment. Tracking compression buys nothing here (you already staged for peaks) and, misconfigured, it pumps your noise floor up during pauses — precisely the artifact QA hunts for.
- Does 24-bit recording change the staging targets?
- It relaxes the danger of recording slightly low — 24-bit files have enormous downward headroom. It changes nothing about noise: your room and preamp hiss don't care about bit depth, and boosting still lifts them identically. Stage the same; enjoy the safety margin as safety, not as license.