I do not sing well.
In my band, the Recalcitrants, I sang lead until some (no-doubt well-meaning) friends took me aside and suggested that we get a lead singer who could actually sing. This was, by any measure, harsh, but I did not take it personally; I stepped down with, I think, quiet dignity.
However, when I write songs for the theater, I have to make demos to let the real singers know how the song goes. But recording my singing is time-consuming and unpleasant. So when I heard that synthesized voices using artificial intelligence (AI) were getting so good that it was hard to tell them from real voices, I was interested.
It turns out there are many companies selling such software. The most common type requires you to send a recording of your own singing to a website and then they will send back a version with another person’s synthesized voice instead of yours. Well, that isn’t helpful. If I could sing I wouldn’t have been looking into this in the first place!
But Synthesizer V, by Dreamtonics, only requires you to enter the notes and the lyrics into your computer and the program will sing. That’s just what I wanted.
And I believe it would make short work of producing demos if I were content with a rather rough demo. But SynthV gives you lots of options to improve the vocals and I haven’t yet been able to resist spending lots of time doing that.
Of the four songs I just added to The Gentlemen’s Ploy page, three used SynthV: “Love Conquers All” has me singing with a chorus of synthesized voices, “Finer Than Silver” uses two synthesized voices (Kevin and Eleanor Forte), and “My Guardian Angel Sleeps” uses just Eleanor Forte.
Interestingly, Eleanor Forte is quite popular on the internet and even has her own fan site, even though she doesn’t actually exist. (From it we learn that she is “honest, endeavoring, and [has] an unwavering sense of morality,” which I personally find so heart-warming.) Kevin, on the other hand, doesn’t even have a last name.
The AI voices do have their limitations. They were designed for contemporary pop music where a large emotional range is not usually required. A greater emotional range is, however, required for musical theater. I used all the tricks I could find to increase the emotions on “My Guardian Angel Sleeps.” Let me know (at phil@prpinnock.com) how you think it turned out.
But why stop with voices? There is software that will write an entire song for you in whatever style you specify, complete with melody, harmony, lead and backup vocals, and instrumental parts. All you do is feed it the lyrics. And I'm sure there are programs that will write lyrics as well.
Of course this has stirred up a cauldron of controversy. Even setting aside that some unscrupulous companies use singers’ voices without their permission, is it ethical to put songwriters, vocalists, and other musicians out of work by using AI? Is it inevitable?
If I were in charge, it wouldn’t be. The first thing I would do is to stop granting copyrights to anything written using AI. Afterall, copyrights were established to protect people—authors, composers—not machines. AI could still be used to write and produce music, and that music could be sold in the millions and make somebody billions, it just couldn’t be copyrighted. That might provide less incentive to use AI, because your catchy little jingle could be used by anyone else, including for purposes you really don’t like. (And without having to pay you royalties!)
But what about AI in the greater world outside of music? There are those who fervently believe that AI will be the salvation of the human race and others who equally as fervently believe that it will be its destruction. I suspect it will be neither. But given that AI is currently being developed primarily as a way to increase the fortunes of wealthy people, I’m sure it will all work out for the best.