Find the mouth click. Turn it down. Fix the noisy gap. Adjust the level. Tweak the compression. Check the loudness. Repeat.
That work still matters, but the way you do it has changed.
Speech-processing algorithms are now good enough to take a lot of the repetitive technical work off your hands. Noise reduction, de-clicking, levelling and mastering no longer have to mean building a complicated plugin chain and watching every meter manually.
That doesn’t mean the whole audiobook should be handed over to software.
For me, the useful distinction is pretty simple:
If the job is technical and repetitive, it’s probably worth automating. If it changes the performance, it probably still needs a person.
The jobs I’m comfortable automating
There are quite a few parts of audiobook post-production that I’d now trust a good automated system to handle.
Mouth clicks and lip smacks
This is a good example of work that can take ages manually without adding anything creative.
If you’ve got dozens or hundreds of small mouth clicks across a chapter, there’s not much artistic value in finding every one yourself.
Automated de-clicking can take care of most of that very effectively now.
You still want to listen back, especially around consonants or unusual mouth sounds, but I don’t think a narrator should have to spend half an hour drawing around tiny clicks if the software can do most of that reliably.
Background noise
Fans, air conditioning, distant traffic, computer noise and steady hiss are all good candidates for automatic reduction.
The important word there is reduction.
You don’t necessarily want to remove every trace of the room. We talked about that in How much noise reduction is too much for audiobook narration? A small amount of room tone can actually help narration feel natural.
The software can do the cleanup. You still need to decide when enough is enough.
Reverb and room sound
Moderate room sound can also be reduced automatically now.
That’s particularly useful for narrators working from home who have a decent setup but not a completely dead vocal booth.
Again, I wouldn’t try to turn every recording into something that sounds like it was recorded in a vacuum.
The job is to bring the room under control so it stops distracting from the voice.
EQ and tonal balance
This is another area where algorithms have become genuinely useful.
A voice can end up slightly boomy, thin, harsh or dull depending on the room, microphone and position.
You can absolutely correct that manually, but if you’re applying broadly similar tonal adjustments across chapter after chapter, there’s no reason the first pass can’t be automated.
I still think this is something worth checking by ear rather than blindly trusting, but the heavy lifting can happen automatically.
Levelling and compression
This is probably one of the biggest time-savers.
Audiobook narration needs to stay comfortable to listen to without becoming completely flat.
That means controlling the big level swings while keeping enough dynamic movement for the performance to still have shape.
Doing that manually over a long chapter can be tedious.
A good automatic process can get you most of the way there much faster.
You just don’t want it pushing every sentence towards exactly the same level.
Loudness and final mastering
This is probably the clearest automation candidate of all.
Once the edit is finished, checking loudness, peaks and the final delivery level is technical work.
There’s no creative reason to manually chase the same numbers chapter after chapter if software can measure and hit them consistently.
The same applies to keeping chapters broadly consistent with one another.
For long-form projects, repeatability matters.
Breaths are where it gets more complicated
Breaths sit right on the line between technical cleanup and performance.
A loud close-mic inhale can absolutely be reduced automatically.
A quiet breath between sentences might be completely fine.
A gasp before an emotional line might be essential to the performance.
That’s why I don’t think breath reduction should ever be treated as a completely automatic decision.
Optivox has Off, Subtle and Max breath settings for exactly that reason.
Subtle reduces detected breaths by 6 dB, which is often enough to make them less distracting without removing them. Max goes much further. Off leaves them untouched.
But whichever setting you use, I’d still listen through afterwards.
If you want more detail on that, the separate guide Breaths in audiobook narration: remove them, reduce them or leave them alone? goes into the different approaches.
The things I would keep manual
This is the part where automation reaches its limit.
Pacing
Software can measure pauses.
It doesn’t know whether a pause feels right.
The gap before a reveal, the silence after an emotional line, the speed of a conversation between characters, those are performance choices.
I wouldn’t want an automatic system tightening or stretching those moments without me deciding.
Gaps between sentences
This is related to pacing, but worth separating because it’s very easy to over-automate.
A gap that looks too long on a waveform might sound perfect in context.
A perfectly consistent half-second between every sentence might look clean and sound awful.
Timing is part of the storytelling.
Dramatic breaths and gasps
A system may hear a gasp and categorise it as a breath.
That doesn’t mean it should be removed.
If the character is frightened, exhausted, shocked or deliberately hesitating, that sound belongs to the performance.
This is exactly the sort of thing that still needs a human ear.
Misreads
Automation can detect speech.
It cannot reliably know whether you said the wrong word from the manuscript.
That’s still a proofing problem.
Same with skipped words, repeated lines, wrong names and pronunciation mistakes.
Performance choices
This is the biggest one.
Software shouldn’t be deciding which take feels better.
It shouldn’t decide whether a character sounds angry enough, whether a sentence needs more space, or whether you should keep an imperfect but emotionally stronger read.
Those decisions are the actual creative work.
That’s the bit I’d protect.
Automation is most useful when it removes repetition
This is where I’ve noticed the biggest difference in my own work.
With podcasts, I used to spend roughly 30 minutes per episode doing the technical finishing pass.
I’d tweak the levels, adjust compression, clean up the audio and make sure the final mix sat where I wanted it.
Now I can drop the full unmixed episode into Optivox and have that part of the job finished in around two minutes.
That doesn’t mean the whole podcast is automated.
I still make the editorial decisions. I still decide what stays in, what gets cut and how the show should feel.
I’ve just stopped spending time on the part of the process that is basically the same every week.
That’s how I think audiobook automation should work too.
Don’t automate judgement just because you can automate processing
There’s a temptation with new audio tools to ask how much of the workflow can be automated.
I think the better question is which parts are worth automating.
Removing clicks? Absolutely.
Reducing steady background noise? Usually.
Levelling a chapter? Yes.
Choosing whether an inhale adds something to the performance? Probably not.
Deciding how long a dramatic pause should last? Definitely not.
The technology is at its best when it handles the mechanical work and then gets out of the way.
Where Optivox fits
That’s also the line I wanted Optivox to sit on.
It handles things like background noise, room sound, mouth clicks, tonal balance, levelling, compression and mastering.
It can reduce breaths too, if you want it to.
What it doesn’t try to do is choose your takes, rewrite your pacing or make editorial decisions about the performance.
Those are still your decisions.
That split makes sense to me because the technical work is where automation can save a huge amount of time without changing what the audiobook actually is.
The whole point isn’t to automate the narrator.
It’s to stop the narrator spending hours doing jobs that no longer need to take hours.