The ear as editor: how I choose among dozens of versions

Old publishing houses had a job I find fascinating: the proofreader. They didn't write the book or sign it. They received the galleys, put on their glasses and read slowly, hunting for the typo, the widow line, the word that had slipped in twice. If they did their job well, nobody remembered them. If they did it badly, the mistake went to print in thousands of copies.

With my songs I do something similar, only with my ears. I write the lyrics, and when I take them to the artificial intelligence tool I use to produce the music, a single set of lyrics turns into many versions. Sometimes a handful, sometimes dozens. Each with its own melody, its own rhythm, its own way of phrasing the lines. And out of all of them I have to keep one. I've already written about what discarding means and the fear of throwing away the good one. Here I want to talk about something else: how I listen, what my ear is searching for, and which flaws make me skip to the next version without a second thought.

The dictation test

I do the first listen with the lyrics in front of me, the way you'd follow a score. But the second one, if the version has survived, I do without looking. I call it the dictation test: could I write the lyrics down on a sheet of paper using only what I hear?

It sounds like something a strict schoolteacher would demand, and it is. My songs live on what they say. If a joke can't be understood, it isn't a joke; it's noise with good intentions. If an ex's name gets smudged, the revenge loses its aim. There are gorgeous versions, with a melody that wraps around you like a blanket, where the words melt into each other and the line arrives as a murmur. In other kinds of music that can be a virtue. In mine it's almost always a problem.

That doesn't mean everything has to be enunciated like the evening news. A song isn't an elocution exam. What I ask is that the words carrying the song, the title, the closing rhyme, the punchline, arrive whole. The rest can breathe.

It's a way of reminding myself what the music is for in my songs. I love it when a production is beautiful, but beauty comes second. The music has to serve the lyrics, the way a good dress serves the person wearing it. If you notice the dress more than the person, something has gone wrong.

Where the weight falls

Spanish has a very particular feature: every word carries its stress on one syllable, and not just any syllable. «Cateto» has it in the middle, so does «Albacete», «reclamación» has it at the end. Poets have been counting syllables for centuries because of this, and the great Spanish copla lyricists knew it by heart even if they never put it in technical terms. The way the stresses of the language line up with the stresses of the music has a name: prosody.

When prosody works, you don't notice it. The line sounds natural, as if someone were really saying it. When it fails, the ear stumbles even if it can't explain why. It's the feeling that a word has come out twisted, like hearing "baNAna" turned into "BAnana" because the melody needed a beat where the language doesn't have one.

It's the mistake that makes me discard more versions than any other. A version can have everything going for it, the rhythm, the atmosphere, a chorus that lifts, and still die because of a single badly stressed word in the most visible spot. Because if the twisted word happens to be the title, it'll be repeated eight times, and by the third you no longer hear it as a word but as a defect.

Questions are similar. In «Tan Cateto» the chorus is a whole question:

Pero ¿cómo puede ser usted tan cateto?

Roughly: "But how can you be such a bumpkin?" That line has to sound like a question, and not just any question: the kind you ask with one eyebrow raised, using the formal «usted» of someone keeping their distance without losing their manners. If the melody flattens it into a statement, the irony vanishes. And a version without the irony of that line is a different song, not the one I wrote.

Strange pronunciations

Then there are the words the machine isn't quite sure how to say. My lyrics have plenty of them, because I write the way people talk on the street and on their phones: app names, anglicisms, brands, acronyms, proper names. In «Mi novio de Albacete» there's a line that goes «Te ha faltado tiempo para poner en Wapo», more or less "you couldn't wait to get on Wapo", and a dating app isn't in any dictionary. Every one of those words is a little lottery.

Sometimes the lottery pays off and the pronunciation is exactly that of a guy from Lavapiés telling his best friend the gossip. Other times the word comes out with an accent from nowhere, neither here nor there, and the line floats around without an owner. And sometimes a letter appears that wasn't there, or one disappears that was, and the line means something else entirely. Given how explicit some of my lyrics are, that kind of slip can be hilarious or disastrous, and almost never in the direction I'd like.

Numbers are another slippery patch. In «3 años de garantía» the consumer affairs office replies with its counter-clerk formula:

Recibida su reclamación con número de referencia

"Your complaint has been received, with reference number..." and then comes a string of digits that has to sound exactly like that, like someone reading out a case file without a shred of emotion. There, the version I'm after isn't the most melodic one, but the one brave enough to be a bit boring on purpose. The joke is in the bureaucratic indifference, and an overly enthusiastic melody kills it.

I've already written a whole piece about regional accents, the Latin American seseo and rhymes that change from country to country, so here I'll only note one thing: I'm not trying to make everything sound like Madrid. I'm trying to make nothing sound like a mistake.

The chorus that won't take off

If I had to choose one word for what I expect from a chorus, it would be take-off. The verse is the runway, the plane taxiing, picking up speed. The chorus is the moment the wheels leave the ground. You have to feel a change in altitude, even a small one.

In some versions that never happens. The verse is fine, the production is fine, and when the chorus arrives everything stays the same, at the same height, with the same energy. The title shows up and nobody greets it. It's like someone walking into a party and nobody looking up from their drink. I drop those versions early, because I know they don't fix themselves: a chorus that doesn't take off on the first listen won't take off on the tenth.

There are also choruses that take off in the wrong direction. They turn solemn when the lyric is a joke, or go light when the lyric is saying something serious. In the chorus of «Mi novio de Albacete» there's an outburst right in the middle:

¡Tu ojete!

It's a cheeky Spanish exclamation, something like "your arse!", and it needs to land like a slammed door, with a breath of air before it so it hits on time. If the melody drags it along without a pause, the shout gets diluted and what should be a belly laugh becomes a mumble. In that chorus, the half second of silence before it does as much work as the word itself.

The atmosphere

There's a question I ask myself even before I get to the details: where is this song? I don't mean the style, which I decide beforehand. I mean the room the sound draws. Some versions sound like a kitchen with the radio on, others like a dance floor at six in the morning, others like a village fiesta under paper lanterns, others like an empty flat on a Sunday afternoon.

The lyrics usually know where they want to be. A heartbreak song with a sense of humour often asks for a bit of party, because the fun lies in dancing on top of the disaster. A complaint to the consumer office asks for something everyday, like running an errand. When a version's atmosphere doesn't match the lyric's, the song looks like an actor who's walked onto the wrong set: delivering the lines perfectly, but in another play.

Sometimes the contrast works, and I go looking for it on purpose. A painful lyric over a beat that makes you want to dance can say more than a painful lyric over a sad piano, because that's how pain is often lived: on the dance floor, drink in hand, putting on a brave face. But that contrast has to be chosen. The difference between irony and a wrong tone is that irony is clearly deliberate.

When the ear gets tired

This is the part people talk about least and the one that took me longest to learn. The ear wears out. Not the way a muscle does, which you can feel, but more treacherously: you keep listening, you keep having opinions, but your judgement is no longer reliable.

Musicians know a phenomenon called demoitis: falling so in love with the first demo of a song that every later version sounds worse simply because it's different. It happens to me with versions. The first one I like becomes, without my deciding it, the yardstick. I listen to all the others against that one, and the more I hear it, the more right it seems, even when it isn't. The ear mistakes the familiar for the good.

Over time I've set myself a few rules to defend against that. They're not foolproof, but they help:

• Stop before everything sounds the same. When I notice I can no longer really tell one version from another, I switch off and walk away. Carrying on in that state is the quickest way to choose badly.
• Come back the next day. A rested ear hears things a tired one skips. Sometimes the version that had me won over at night sounds limp in the morning, and the one I'd left in the second row suddenly steps forward.
• Listen at low volume. At high volume everything impresses. At low volume only what's well built survives: the melody you can follow, the key word you can hear, the chorus you recognise even in the background.
• Listen while doing something else. Washing up, sorting papers, answering a message. If the chorus makes me look up even when I'm distracted, it has something. If it slips past without my noticing, it doesn't.

None of these rules are mine alone. They're tricks of the trade that existed long before artificial intelligence. What has changed is the quantity: when you have many versions of every lyric and some 273 songs released, a tired ear stops being an anecdote and becomes part of the job that has to be organised.

The final decision

In the end, among the versions that have passed every filter, I have to pick one. And I've realised I almost never choose the most spectacular. I choose the one that never makes me wince.

That's the ultimate test for me: listening to the whole song from start to finish and checking there isn't a single second where I think "oops". No twisted word, no chorus stuck on the ground, no change of atmosphere that comes out of nowhere. Another version might have a more brilliant moment. But if it also has an "oops", it loses. I'd rather have a well-rounded song than a song with one stroke of genius and one stumble.

Once the decision is made, I stop comparing. That took me a long time, because the temptation to go back and listen to the rejected one one last time is enormous. But an editor who never lets go of the manuscript never publishes the book. Once chosen, the version becomes the song, and the rest goes into the archive.

I don't know whether my ear is better or worse than anyone else's. I know it's the one I've got, that I've trained it by listening a lot and getting things wrong quite often, and that it's the part of my work nobody sees but that you can hear in every song. Like the proofreader, if I do my job well, nobody will notice I was there.

If you want to check the result, all the lyrics are on the lyrics page and the songs are free to download in downloads. They're also on Spotify and Deezer.