Singing robots – Autotuning and an ode to Imperfection

Some of you will know what an autotuner is. Others will have heard it on songs but not known quite what is going on. Others out there will be completely oblivious to it when listening to music. After reading this, you should hopefully all be able to hear this in radio friendly pop and should more importantly have some idea of what it means to music in general.

Autotuner is an effect producers can use on vocals (or other instruments; but mainly vocals) to “bend” notes into tune. The way it works is fairly simple. Autotuner analyses the frequency of the note sung, also known as the pitch, and references that against the key of the music. It can then correct flat or sharp notes by mechanically changing the pitch to the right note. For instance: Middle A on a piano is 440hz. If the singer sings 438hz, it will sound flat. The autotuner knows this and will bend it up to 440hz, placing it in tune.

Fantastic! Or so it would seem. Let us delve deeper.

There are myriad issues for music alongside the benefits. Let’s start with the sound itself

Autotuners, when used on high settings, do not sound natural. That “inhuman” sound that pop singers have nowadays, where everything is very polished, where all notes are perfect and there is no movement between them? Most likely an autotuner set high. It sounds robotic, because it is. When a “human” person sings, they move between notes. There will be a small degree of sliding, especially if the notes are held as they move between pitches. This is normal. Removing these sounds odd, especially if the removal is harsh and sudden.

There are two ways autotuner can be used well in certain circumstances. It can be used, very carefully, to tidy a good but not perfect vocal performance. Autotuner has two main edit buttons –  “speed” and “choosiness”. Speed is how fast the note is bent into pitch from the incorrect pitch, and choosiness is how much leniency the tuner allows (i.e. if aiming for middle ‘A’ at 440hz, allow a few hertz either side). Using it at a reduced level on both will tidy up a slightly ropey vocal without sounding too robotic. However, a producer’s ear and those of a general music aficionado are rarely fooled. And again, the “something is not quite right here” effect will hit even the layman if it is set even a tiny bit too high.

Secondly, autotuner can be used outrageously, explicitly, as an “effect”. Creative dance music does this. It was first known as the “Cher effect” in the nineties for its use on the track “Believe”. The robotic sections where she sounds like a machine singing are the result of autotuner set on maximum. No one listening to that record thinks she sang it that way naturally, it is clearly digital trickery. As such, it is an effect. And this is fine, brilliant even, if done well. Creative tools are good. The key word here is “creative”.

(An interesting side note: The small London studio that produced “Believe” for Cher were so keen on keeping this creative abuse of autotuner a secret, they actually lied for a period of time to the press, claiming a different effect was employed with bespoke settings. The scamps!)

However it seems more and more, autotuner is used to cover bad singing.

Since it has become readily available, it would seem nearly anyone can have a radio friendly record. Granted, the vocalist needs to provide at least one take that is close to being in tune, but most anyone except the most tone deaf can manage this! As such, guess how pop singers are chosen; marketability and/or sex appeal.

Yes. I know. Pop stars have been sex symbols since Lennon and McCartney. And before that Elvis. And umpteen other examples pre-digital recording are available. But these were pop stars that had talent – huge, world conquering talent – as well as good looks. Now, such a thing is almost a moot point. A singers saleability is seemingly determined by looks over actual singing in the pop market.

The game is up when singing live, one would assume. Alas, no. CPU speeds are now sufficiently zippy so that a vocal can now be processed on-the-fly, with only the tiniest lag from microphone to loudspeaker. The show ‘X-Factor’ came under fire last year for using Autotuner live on stage, and rightly so. This competition is supposedly about performance prowess, not computational trickery.

In the highest reaches of the pop charts, Autotuner is most dangerous. Let us examine some of the artists du jour.

(Note; this is not an attack at either of these artists. They are doing very well for themselves, they have some catchy-as-hell singles and bring joy to many people. I am simply focusing on one factor and how is effects their individual musicality).

Katy Perry. Her debut was a quaint pop record with some clever hooks and a human edge. “I kissed a girl” broke her, and the vocal is tidy but not over processed. Most importantly is was a good, catchy pop song. By album two, the quaint has gone (along with apparently every ounce of body fat?!) and she instead settles for autotunery of the most callous kind. Simple loops and over processed vocals with no substance to be seen. Yet she sells more than ever…

Rihanna. Again, her debut record was poppy, fluffy but human. And she is now so robotic on record – and often quite notably off key live – as to be nearly unrecognisable when these two facets are placed side by side

Parallels with the “look” of these artists can also be drawn, so perhaps there is a deeper factor here. Both these artists are now super-skinny Barbie dolls. The superficial has overtaken them. Maybe the autotuned quote-unquote perfection of their records is part of this sheen, a perfection beyond that a human can provide. But this is false, quasi beauty. Removing the makeup, the hair, the autotuner and the production trickery, and beneath there is likely very little natural beauty.

In the same way that photographic studies have proven that digital replicas of human faces with an exact line of symmetry look odd, inhuman and ultimately unattractive, I find such heavily processed vocals do not have the same appeal to our humanity as normal, well sung music. Perfection is not ideal. A vocalist should of course aim to be in key, to hit to notes. But the movements between, the slight vibratos, the bends and ticks all provide character. Nuance.

If nuance and colour were of no concern, all instruments would have been replaced by the synthesiser 20 years ago. They haven’t. Guitar, strings, vocals et al continue and always will.

Don’t misread me. I love a good synth song. And in the right setting I love an awesome autotuned, robotic vocal. Daft Punk brought these aspects together on their seminal “Discovery” album which is one of the most played albums in my collection.

But let’s use the tools we have carefully and creatively. Let’s leave some of the musician in the music.