Recording an audiobook at home is one of the few audio productions where technical quality is measured objectively. ACX (Audible/Amazon’s audiobook platform) has published specifications: maximum background noise floor, acceptable loudness range, file format. Either you pass the specs or your file gets rejected at submission.
The good news: those specs are achievable with a home setup. The complicated part isn’t the acoustics — it’s recording 6, 8, 10 hours of content across days or weeks and keeping your timbre consistent from chapter 1 to chapter 40.
That’s where the voice changer comes in — but let’s build the foundation first.
What ACX Actually Requires
The ACX technical specifications that reject the most submissions are:
- Loudness: between -23 and -18 LUFS RMS, peak maximum at -3 dBFS
- Background noise: below -60 dBRMS in the silence between speech
- Format: MP3 192kbps or WAV 44.1kHz/16bit, mono or stereo
Background noise is what fails the most people. It’s not a bad microphone — it’s the AC unit you’ve stopped noticing after an hour of recording, or the refrigerator that kicks on exactly when you’re mid-paragraph on the most beautiful part of the chapter.
Minimum Setup That Works
Microphone: dynamic, not condenser. The Samson Q2U (USB + XLR, around $50) is the entry point most independent authors use. If you have more budget, the Shure MV7 handles it. A dynamic mic has a cardioid pattern that rejects lateral and rear noise — in a home environment, that’s worth more than the “superior quality” of a large-diaphragm condenser in an untreated room.
Acoustic treatment: you don’t need a booth. You need a room with:
- Soft surfaces (clothes, curtains, carpet, a mattress nearby)
- No large parallel bare surfaces (wall + floor = reverb that will appear in the recording)
The cheapest trick: record inside a closet full of clothes. Sounds like a joke, but it works better than a lot of expensive solutions.
Audio interface: if your mic has XLR output, use a simple interface (Focusrite Scarlett Solo). If it’s USB direct, that’s fine — but avoid long cables without ferrite to reduce line noise.
Timbre Consistency Is the Real Problem
Here’s the challenge that most technical guides completely ignore.
You start the audiobook in January, rested, full voice. You record two chapters a week. In March, you had a cold. In May, the seasons changed and the air got dry. By October you’ve finally reached chapter 38 — and your voice sounds different from chapter 1.
A listener who binge-listens the audiobook will notice. They won’t be able to articulate the problem, but they’ll feel like “the voice changed” partway through the book.
VoxBooster solves this as a standardization layer: you record your raw voice normally, then process each session with the same clone profile. The model preserves your performance (rhythm, emotion, intent) and normalizes the timbre to the same reference point. The chapter recorded on a bad day will sound consistent with the chapter recorded at your peak.
The flow: record the chapter, export, drag it into VoxBooster in offline processing mode, apply the chosen profile. Out comes a file with standardized timbre. Only then do you do loudness and noise treatment.
Recording Flow Per Chapter
- Warm up your voice before recording — 5 minutes of reading aloud before hitting record makes an audible difference on the first paragraph
- Record the whole chapter, not paragraph by paragraph — frequent interruptions create performance inconsistency; if you make a mistake, say “ERROR” aloud and repeat from that sentence, don’t cut in the moment
- Leave 2 seconds of silence before and after each take — you’ll need this to measure background noise floor in editing
- Export as WAV before editing — never edit the final MP3; keep the WAV as your master file
- Process in VoxBooster with your standardized voice profile
- Edit in Audacity or Reaper — remove marked errors, clean up heavy breaths, apply noise reduction if needed
- Normalize to -19 LUFS (center of ACX range) with limiter at -3 dBFS
- Measure background noise — select 1 second of silence and check the RMS; if it’s above -60 dB, apply a noise gate or noise reduction before normalizing
On Narrating Fiction vs Non-Fiction
For technical non-fiction, a more neutral and consistent voice works better — the listener wants clarity, not dramatic performance.
For fiction, you have more latitude. If the book has dialogue, it’s worth using slightly different voice profiles for main characters — not enough to turn it into an audio drama, but enough for the listener to distinguish the narrator from the protagonist when they speak in first person.
VoxBooster lets you save separate profiles: “standard narrator” as the base and “protagonist” with a slight timbre variation. This adds a layer of production that independent audiobooks rarely have.
Before Submitting to ACX
Run the final file through the ACX Check plugin (free for Audacity). It measures loudness, peak, and background noise and tells you exactly what’s out of spec. Don’t submit without passing this check — rejection delays you by weeks and the resubmission process is bureaucratic.
Recording audiobooks at home is slow, requires discipline and process consistency. But it’s completely viable, and the distribution platform is the same one the big studios use.