You can get a recording technically cleaner and still make it sound worse. Push noise reduction, compression and other processing too far and the narrator starts to lose some of the movement that made the performance feel natural in the first place.
Audiobooks are especially sensitive to this because you're not listening to the voice for 30 seconds. You might be listening to it for eight or ten hours.
The aim isn't complete silence behind the narrator. It's getting rid of the things that distract you without taking the life out of the recording.
An audiobook shouldn't sit at one volume for ten hours
One of the easiest ways to make narration feel over-processed is to flatten the dynamics too much.
When I'm listening to an audiobook, I want to rise and fall with the tension of the performance. A quieter passage should actually feel quieter. When the narrator leans into a line or raises their voice, I want to feel that too.
Obviously, there is a limit.
If the quiet sections disappear into the background and the loud sections have you reaching for your headphone volume, there is too much variation. Good mastering should control that difference.
But controlling the dynamics isn't the same as removing them.
That's quite different from some other types of spoken-word audio. On a podcast, for example, having the host and guest sit at a very consistent level is normally desirable. A commercial voiceover is often meant to sound very controlled and present as well.
With an audiobook, some of those dips and rises are part of the performance.
If everything ends up pushed towards exactly the same level, you can still hear that the narrator is acting out a tense or intimate moment, but the audio is no longer helping to sell it.
Don't turn the room into digital silence
Another giveaway is what happens in the gaps between phrases.
Even in a good recording space, there is some room tone. A quiet booth isn't literally silent.
If you process a recording aggressively enough that all of that disappears whenever the narrator stops speaking, the gaps can start to sound unnatural. You hear the room while they're talking, then suddenly there is nothing, then the room comes back again with the next sentence.
Once you notice it, it's difficult not to hear.
I'd rather have a very small, consistent amount of room tone sitting underneath the performance than hear the processing opening and closing around every phrase.
That doesn't mean leaving obvious fan noise, traffic, hiss or hum in an audiobook. Those things are distracting and should be dealt with.
It's more that zero background sound isn't necessarily the target.
A little room around the voice can actually help
There is also a difference between making something sound clean and making it sound completely dry.
Commercial voiceovers are often deliberately very close and present. That can work because you're normally hearing them for a short amount of time and the voice needs to cut through immediately.
An audiobook can benefit from having a little more air around it.
Not obvious reverb. You don't want the listener thinking about the room. But stripping every trace of space out of a recording can leave the narrator sounding unnaturally close to your ear.
For long-form listening, I generally prefer something that still feels like a person performing in a real space, just a very controlled one.
That balance will vary by project too.
For romance or fantasy, for example, you might want a slightly more intimate sound. Bringing the narrator a touch closer by reducing a little more room sound or reverb can work really nicely.
With non-fiction, I'd usually favour something simpler and more consistent. Clear, controlled and easy to understand without trying to make the sound itself too artistic.
Neither approach is a rule. They're choices.
Breaths aren't automatically something to fix
Breaths are another good example of where technically cleaner can become less natural.
A quiet inhalation between sentences often doesn't bother me at all. Sometimes it actually helps the rhythm of the narration.
If you remove every one, the spaces between sentences can start to feel oddly empty.
That's different from a huge close-mic inhale that suddenly jumps out at the listener. I'd normally want to reduce something like that because it pulls your attention away from the story.
There are also breaths that are obviously part of the performance. A gasp, a nervous inhale, a character catching their breath. Those are editorial decisions, not technical problems.
This is why I tend to think of breath processing as reduction rather than automatic removal.
We've gone much deeper on this in Breaths in audiobook narration: remove them, reduce them or leave them alone?, because it's one of those areas where there isn't a single correct answer.
Some narrators genuinely prefer almost all breaths removed. Others want them left largely untouched. A lot sit somewhere in the middle.
Noise reduction can become more distracting than the noise
There's a point with most cleanup tools where you start trading one problem for another.
Keep pushing noise reduction and you might begin to hear the voice thinning out, developing metallic artefacts or changing slightly around the ends of words.
At that point, I'd rather hear a small amount of steady background noise.
The listener's brain is surprisingly good at ignoring something consistent. A quiet noise floor tends to disappear once you're absorbed in the book.
Processing artefacts are different because they move with the speech. They keep changing, so your ear notices them.
That's normally my cue to back the processing off.
The same goes for compression
Compression isn't noise reduction, but it creates a similar problem when it's overdone.
It's incredibly useful for audiobook mastering because you do need to keep the listener within a comfortable volume range.
But if a whisper ends up nearly as loud as an emphatic line, you've probably gone too far.
I'd describe the target as controlled dynamics rather than flat dynamics.
The listener should still feel the quieter and louder moments. They just shouldn't have to manage those moments themselves with the volume buttons.
That's a surprisingly fine balance.
So how do you know when you've gone too far?
There isn't a useful number I can give you for this because it depends on the recording, the narrator and the book.
I tend to listen for a few things:
- Does the voice start sounding thin, metallic or unnaturally smooth?
- Can I hear the room disappearing and returning between phrases?
- Have quiet and loud moments started feeling too similar?
- Have normal breaths and little performance details disappeared?
- Am I now noticing the noise reduction more than I noticed the original noise?
- Does the narrator still sound like they're performing, rather than simply producing a perfectly even stream of speech?
The last one is probably the most important.
Sometimes the best way to judge a processed recording is to stop listening for technical problems altogether.
Listen to a few minutes as if you'd bought the audiobook.
Are you following the narrator and the story, or are you noticing the audio?
If you're thinking about the processing, something probably needs adjusting.
Clean the recording without cleaning out the performance
This is a big part of the thinking behind Optivox.
There are plenty of technical jobs that can now be automated very effectively: noise reduction, reverb reduction, tonal balancing, levelling and the final mastering pass.
If you'd like to hear that balance on your own narration, Optivox handles the full audiobook mastering pass with exactly this in mind.
But the aim isn't to make every voice perfectly uniform.
An audiobook still needs movement. It needs space. Sometimes it needs breaths. It needs quieter moments to feel quieter so the louder ones actually mean something.
The best processing is normally the processing you stop noticing.
You hear the narrator.
You hear the story.
And after a while, you forget about the audio completely.