File Formats and Bit Depths – WAV, MP3, 16-bit and 24-bit
There are a number of different formats for audio files, but the purposes of audiobook production we only need to know about two – WAV and MP3.
MP3 will be the delivery format for sending the audiobook to Audible. WAV is the origination format when you’re recording and editing the material.
Although both formats support different bit depths, the MP3s uploaded to Audible can only be 16-bit. The WAV format you use to record and edit the audiobook can be either 16-bit or 24-bit (or even 32-bit if you’re so inclined but I wouldn’t recommend it).
You might question what, if the delivery format must be 16-bit MP3, is the point of originating in 24-bit WAV? Well, as to the MP3/WAV part of the question, no professional DAW records directly to MP3. And 24-bit? I’ll get into that in a minute.
The difference between WAV and MP3
Both formats are audio files containing those binary numbers we talked about here. But while WAV files contain every binary number that was originally recorded, MP3 files are created using a lossy compression algorithm that saves space by (amongst other things):
- removing very low and very high frequency sounds
- removing very quiet sounds that occur close to very loud sounds
- Numerical rounding of data to simpler numbers requiring fewer bits to store
How much of this happens is a trade-off defined by the encoding bitrate, something that is set by the software (usually the DAW exporting the audio). Bitrates below 192 kbps are not generally accepted by Audible. I export at 256 kbps if I can. (More details on MP3 exports will appear in a later article on exporting for Audible.)
So, every time an audio file is converted from WAV to MP3, a little bit of the information is lost. Repeated conversion between the two formats will degrade the audio, like a bad photocopy.
WAV or BWAV?
By the way, your DAW is almost certainly using BWAV rather than WAV. There’s no difference in the audio content; BWAV has additional embedded metadata like timecode, timestamps and so forth. If you see BWAV, think WAV.
The difference between 16-bit and 24-bit
In this article, I mentioned recording at a lower nominal level to give yourself plenty of headroom. This way of working has only become practical since DAWs went 24-bit.
I’ve often been asked why I bother recording and editing in 24-bit resolution only to have to export the audio in 16-bit to suit Audible? There are several reasons for this, which I recommend you consider yourself.
16-bit “goes” from 0 to 65,535. 24-bit “goes” from 0 to 16,777,215 – it’s 256 times higher resolution.
16-bit doesn’t do well at low volume
In the early days of DAWs, when they all worked 16-bit, we learned to record at as high a level as we could. This is because the lower the volume, the fewer bits are being used. At -18dBFS, a 16-bit system only has 13-bit resolution. At -48dBFS, it’s only got 8-bit resolution (think 1980’s arcade video games). There will be audio at this level – like the s at the end of a word for example. In the 16-bit system, quiet sounds get crunched into a mess of quantisation noise and harmonic distortion.
On the other hand, a 24-bit system still has 21 bit resolution at -18dBFS. At -48dBFS, it’s still got 16 bits in play – equivalent to the entire range of a 16-bit system. That quiet s sound, that gentle outbreath, these will recorded in high fidelity.
Another way of looking at it is that 24-bit recording has a theoretical noise floor below the self-noise of even the quietest microphone.
But what’s point if you have to export to 16-bit anyway?
Being able to record at a lower nominal level gives you more headroom before clipping. But after you’ve edited, tamed the shouting, removed the heavy breathing and so forth you will raise the overall level of the recording as high as it will go and still fit Audible’s requirements – meaning the 16-bit resolution will now be sufficient.
Think about the plugins
I use a lot of plugin processes, including spectral noise reduction and ambiance reduction which run right across the recorded audio, as well compression. limiting and eq, spot noise reduction, spot eq correction and so on. All these processes run internally as 32-bit and then convert down to the session resolution. Repeatedly doing this in a 16-bit session amplifies the problems noted above. It’s that dodgy photocopier again.
The answer is to stay within the 24-bit resolution until the very last moment – and then do one conversion to 16-bit when you finally export to MP3.
Why you need to care
Forgive me if all that got a bit technical. Look at it this way. Recording at 24-bit resolution means you never have to worry about keeping your levels up during recording. You can give yourself lots of headroom for those shouts and laughs with no fear of clipping.
You need to deliver files in 16-bit MP3 format to Audible. But your recording and editing will work better and the finished result will sound better if you originate and edit in 24-bit.
One Last Thing
For many of the same reasons given above, I also originate all my sessions at 88.2 kHz, double the sample rate of audiobooks (or 96 kHz, double the rate of audio-for-video for voiceover)*. You might want to consider this too, although it will increase both your storage space requirements and possibly your computer processor speed requirements.
<<back to room tone – the spaces in-between
on to Instantaneous and average levels: RMS, LU, LUFS>>
*I do try to resist my natural tendency to get over-technical and indulge my love of facts and figures. If there’s something you think I should go into more deeply, drop me a line or write a comment.
Justin Hill has been in the recording business for more years than he likes to talk about but has traded as The Other operation since 2013. In early 2020, just as the COVID lockdowns started, he decided to build a dedicated spoken word studio.
Despite materials and equipment suppliers either being closed or having no stock, and delivery drivers being unwilling to come to the door, Landen Park Studio actually got finished and had its first session in August 2020.
Since then, Landen Park Studio has recorded countless audiobooks and hosted countless voiceover sessions with Justin Hill at the helm. Many of these audiobooks have been self-narrated by authors who’ve never recorded an audiobook before.