Why You Should Never Record Your Voice-over as MP3 and Convert to WAVE

When working with voice overs I always request wave (at least 16bit 44.1kHz) but quite often get a file that started life as an mp3. This is because some voice overs record in mp3 format and then convert into wave.

I have had this conversation many times over the years and I am still surprised at the number of voice overs that don’t understand the perils of recording in mp3 format.

Up-scaling an mp3 to make it a .wav or .aif file does not make it better quality. It just makes it bigger.

Voice overs should always record to wave or aif ideally at 24bit 48kHz or at least 16bit 44.1kHz.

When I say this to voice overs they are shocked that I can tell but when I ask can you send the original wav, I have been informed (on several occasions, this is not an isolated incident) that said voice over(s) always record as mp3. What!

This alarms me as it shows that some professional voiceovers do not understand the difference between lossy and lossless file formats or the implications on sound quality. I can understand that reducing file sizes was good practice back in the day of smaller expensive hard drives and slow internet speeds of the early 1990’s, but there is no reason for it nowadays.

If you are producing any kind of audio it is never good practice to record and produce as an mp3. This format is very lossy meaning that its size is compressed making it smaller by stripping out up to 90% of the audio data.

Simply getting your mp3 master (two words that should never appear together) and converting it into a wave file is not good enough. Once the audio data has been stripped out it can never be put back.

The mp3 algorithm is very clever in that it removes elements of the audio that it thinks you won’t miss. However, the fact is, most of your audio has gone.

Spectral Analysis

This sounds techy and it can be, but it is simply a visual representation of all elements of the audio spectrum.

Firstly, here is a screenshot of a wave that a voice over recorded as 16bit 44.1kHz wave. You will see that the audio data is present all the way from the lowest sound to the highest sound. The scale on the right shows frequencies all the way up to 20kHz. This is how a voice recording should look.

Here is an mp3 converted into a wave file. See that nearly all of the sound above 16kHz has been stripped out. This is the most obvious omission as it can be seen visually, but be aware that there will be much more data missing in the rest of the audio that this spectrograph can’t show visually.

As you can see the data that was not present in the mp3 was not magically returned when it was converted to a wave file. It is gone forever.

Does this really matter?

If you are listening to this on cheap speakers or earbud headphones from your music player then probably not.

However, if you work in audio production or are a voice over it should matter very much.

Your reputation relies on you creating the best quality audio possible and at the very least you should record and produce with lossless audio formats. Essentially anything with the word lossy associated to it should not be used for pro audio production.

I look at it this way, the mp3 converted to wav has been stripped of most of its audio data, so for example, when working on a voice over recording I will carry out several digital processes to polish the sound and may incorporate sound effects or music too. Now, if I denoise the voice, declick it and remove any popping this is all happening in the digital realm and I am missing most of the audio data that should have been recorded. As such, my sound is going to be thinner and  the result I can achieve with good production and processing is going to be seriously affected.

This can make a mockery of you as an audio producer as you may experience any or all of the following:

• Someone notices that the audio originated from a lossy format using a spectrograph and marks you down as an amateur. Especially frustrating if you thought you were working with a wav, but your voice over sent you a wav that started life as an mp3.
• The audio master sounds warbled or watery due to the data compression.
• The audio just sounds bad no matter how much you try to use fancy production techniques to improve it.

What if my client requires a lossy format?

That is quite common and lossy formats are perfect for streaming purposes and online video. However, best practice means that you should always convert down from lossless to lossy.

Converting your studio wave into mp3 for your client, if that is what they want, is perfectly acceptable but always have lossless masters. Your mp3 version should be your inferior copy, not your only copy.