How to master an audiobook for ACX

“Mastering” an audiobook means taking your edited narration and making it sound finished and consistent: clean, even, at the right loudness, and inside the technical limits the platform checks. For ACX those limits are an RMS of −23 to −18 dB, peaks no higher than −3 dB, and a noise floor no higher than −60 dB (our ACX requirements guide explains each one).

The order matters more than any single setting. Each step below changes what the next one sees, so work through them in this order.

Before you master: get it right at the microphone

Mastering can polish a good recording. It can’t rescue a noisy one. Three things at the recording stage save hours later:

• A quiet room. Turn off fans, air conditioning and anything that hums while you record. Soft furnishings, a closet full of clothes or a few blankets do more than expensive foam.
• Sensible gain. Set the microphone so your louder lines peak somewhere around −12 to −6 dB. Too low, and mastering has to boost the noise along with your voice. Too high, and loud moments clip, which nothing can undo.
• Consistent position. Keep the same distance from the microphone, session after session, so every chapter starts from the same sound. A pop filter keeps P and B sounds from thumping.

Record in WAV, ideally 24-bit, and record room tone at the start of every session.

Step 1: Edit first, master last

Do all your editing on the raw recording: remove mistakes and retakes, tighten long pauses, fix pickups. Mastering settings react to the whole file, so if you edit afterwards you change what they measured.

Step 2: Clean up

• High-pass filter around 80 Hz. It removes rumble from traffic, footsteps and handling noise below the voice, without touching the voice itself.
• Plosives, mouth clicks and breaths. Soften popped P and B sounds, remove clicks and smacks, and turn breaths down rather than cutting them out. A narrator who never breathes sounds unnatural, and cut-out breaths leave odd gaps.
• Noise reduction, gently. Learn the noise from your room tone and reduce it by a modest amount. Heavy noise reduction makes a voice sound watery or robotic, which listeners notice far more than a little room sound.

Step 3: Shape the tone with EQ

Use EQ lightly. Most voices benefit from a small cut where the room adds boominess (often somewhere between 200 and 400 Hz) and perhaps a gentle lift for clarity. Make small changes, a few dB at most, and compare with the EQ off. If you’re making big moves, the problem is usually the room or the microphone position.

Step 4: Even out the reading

A natural reading rises and falls: whispered asides, raised voices, the ends of sentences trailing off. A compressor or a speech leveler brings the quiet and loud parts closer together, so the listener doesn’t have to keep adjusting the volume. Gentle settings sound natural: a ratio around 2:1 to 3:1 that turns down only the louder moments by a few dB. A de-esser tames sharp S sounds, which compression tends to make more noticeable.

Step 5: Set the loudness

Now raise or lower the whole file so its RMS sits around −20 dB, the middle of ACX’s −23 to −18 window. Measure the whole chapter, not a sample of it. Aiming for the middle gives quiet and loud chapters room on both sides.

Step 6: Limit the peaks

The last processing step is a limiter with its ceiling a little under ACX’s −3 dB, for example −3.5 dB. It catches the few peaks that the loudness step pushed too high. The margin matters: encoding to MP3 can raise peaks slightly, and the MP3 is what ACX measures.

Step 7: Room tone and the noise floor

Make sure each file starts and ends with 1 to 5 seconds of room tone, and that any gaps you edited are filled with room tone rather than silence. Then measure the noise floor in the pauses. If it’s above −60 dB, go back to the source: the room, the gain, or a little more noise reduction. Don’t gate the pauses down to complete silence.

Step 8: Export and check every file

Export each chapter as a 192 kbps, constant bit rate, 44.1 kHz MP3, all mono or all stereo. Then check every file’s RMS, peak and noise floor, not just the first one: the chapter that fails is always the one you didn’t check. And listen through on headphones for clicks, breaths and missing words, which no meter will find.