Version 31 sounds excessive. Sometimes it is. But making a song isn’t a race to see how quickly I can get from an idea to something listenable.
When I made Song Director, the version I eventually chose was number 31.
That doesn’t mean I spent thirty attempts repairing a bad song. Quite the opposite. Several earlier versions were perfectly respectable. Some had a voice I liked. Others had a particularly good groove or an arrangement that almost worked.
Almost is a dangerous word.
With every new version, I’m listening for different things. Does the voice sound like the person who should be telling this story? Is the tempo helping the words, or rushing them? Does an instrument add character or merely occupy space? Does the chorus actually lift? Does the arrangement know when to get out of the way?
And sometimes the problem is much smaller. A phrase lands awkwardly. A word is stressed incorrectly. The energy drops in the wrong place. The ending goes on too long.
So I change something and try again.
The interesting part is that progress isn’t linear. Version 17 isn’t necessarily better than version 12. A change that solves one problem can create another. A different voice can suddenly make a line work beautifully while making the rest of the song less convincing.
Occasionally I go backwards.
That is also why I don’t particularly like the idea of simply generating lots of versions and picking the best one. The process is more deliberate than that. Each round tells me something about the song. I learn what it tolerates, what it needs and what I should stop trying to force upon it.
Song Director gradually found its identity as a jazzy swing song. By version 31, the pieces finally belonged together.
There is another reason I keep going: technical perfection isn’t really the objective.
A slightly imperfect voice with personality can beat a flawless one. A sparse arrangement can be stronger than an impressive one. One unexpected musical detail can give a song more identity than another layer of polish.
The final decision is therefore surprisingly unscientific.
I listen.
And at some point I notice that I’m no longer making a mental list of things I want to change. I’m listening to the song instead of evaluating the production.
That’s usually the moment.
Thirty-one versions isn’t a target. Sometimes a song arrives remarkably quickly. Others stubbornly refuse to reveal what they want to be.
The number doesn’t matter.
Knowing when to make version 32 – and when not to – does.
