Where I stand on AI and singers

I build a tool that puts a vocal on a producer's track. So I am not a neutral party in this discussion, and you should weigh what I say with that in mind.

Here is what I think anyway, starting with the part that does not help me.

Start with the part that is not comfortable

At the bottom end of the demo market, work is going away.

Not all of it and not everywhere. But there is a kind of job that used to exist and is getting rarer. The quick demo vocal. The scratch take. The guide someone sang for a fee because the producer could not sing it himself. Some of that work is not coming back.

This is not the first time. When the drum machine arrived in the eighties, session drummers lost work. That happened. People who had a steady gig found out that the gig had stopped existing.

What came after was real too. New genres appeared. But they appeared later, and they appeared for other people. If you lost your regular work in 1983, the news that house music was coming would have been thin comfort.

I am not going to tell you that did not happen. And I am not going to tell you that this time it is different for everyone, because for some people it will not be.

So why did I build one anyway

Four reasons. In the order that I actually weigh them.

1. A lot of this work was never paid work in the first place

“Producers don't need singers anymore” is partly true and mostly too short.

Not every producer making a track with Vokals today would otherwise have gone looking for a singer. Many of them had no budget for one. They were never going to hire anybody. The track was going to stay a loop on a hard drive.

That money was not being spent. What is happening there is not work moving away from singers. It is tracks getting finished that were not getting finished.

2. Ideation. This is the one I weigh heaviest.

A producer who does have singers he works with can now decide his own direction first.

He writes the lyrics. He hears the song with a vocal on it, in his own words, before he calls anyone. Then he sends that demo to a real singer and asks her to sing it again.

What she gets is a demo with a worked out topline instead of an instrumental and a rough description of what he had in mind.

This is the one route where the singer comes out ahead. She is not being asked to guess at something that only exists in someone else's head. She is being handed a direction and asked to do it properly.

3. The drum machine

When the drum machine came out, people said drummers would not be needed.

What happened instead is that entire genres appeared that did not exist yet. House. Techno. Hip hop. All of them built on a machine that was supposed to end a job.

That is the pattern with most new studio gear. Something gets automated, and something else gets built on top of it that nobody saw coming.

I do not know what gets built on top of this one. Neither does anyone else, and you should be careful with people who tell you they do.

4. Live is still people

When a project gets big enough and it goes on stage, people go looking for real artists again.

I have not seen that change and I do not expect it to.

Singing and songwriting are two different jobs

This is where most of these discussions go wrong.

Singing is a craft. Songwriting is a different craft. A session singer is not automatically a good songwriter, and plenty of good songwriters cannot sing at all.

What Vokals helps with is the writing end. Hearing whether the idea holds before anyone commits time to it. That is a different job from the one a singer does. Not a smaller one and not a bigger one. A different one.

Most of the heat in this argument comes from treating them as the same job, and then concluding that one is taking the other one's place.

What that looks like in practice

The part that decides whether this works is not the tool. It is how you hand the demo over.

Here is how I do it:

“This is the direction I want to go in. Do you hear anything in here you'd improve?”

That is an invitation and not an instruction. She is being asked for her judgement, not for her execution.

She might tell me to forget the demo and send her the instrumental instead. That happens, and it is a good answer. It is still a better conversation than the one that starts with two reference tracks and a vague description.

Where this comes from

Two things, and they happened years apart.

The first one is my own work. I have been making records for over 30 years. Making the music was never the problem. Getting a voice onto it was.

The demo that became Kate Ryan's “Scream for More” sat there, finished, without a vocal.

For weeks. I had the network. I had the credits. It still sat there. Eventually a singer came in and did it.

The part that matters more than the waiting is what I did while it waited. Like any producer who wants to keep moving, I just built new tracks in the meantime.

That is the thing nobody warns you about. The production never stops. The finishing stalls. And the pile of unfinished work grows.

There are still dozens of demos on my hard drive that never came out. That was true of every producer I knew. It probably sounds familiar.

The second one is not about me. From 2020 to 2024 I ran livestreams where producers sent in their tracks and I gave feedback. Across all of those years one question kept coming back, and it was always the same one:

“Where can I find vocals?”

It did not go away when the world opened back up. It was still coming in years later.

I had two answers. Go to SoundBetter. Or post a call-out on social media.

Neither of them worked very often.

SoundBetter was too expensive for most of them. Posting publicly meant admitting you cannot do something yourself, and a lot of people would rather not. Some were afraid nobody would answer. Some did not want to negotiate with strangers. And a lot of them felt that as an unknown producer they had nothing to offer that would make a singer say yes.

Then there is the one that has nothing to do with money. They were afraid of putting a lot of work into a singer who would not deliver. That is not a price problem. You can pay and still end up with the wrong person, and then you have lost the money and the weeks.

I gave that advice for years and watched it not help.

“It was difficult for me, not being able to help them with that.”

What I was actually trying to build

I did not start this as a company. When vibe coding came along I wanted to see how far I would get on my own, and I was building it for myself. It worked better than I expected, and at some point it stopped being a private tool.

“Mostly, I wanted to build something I would have wanted myself as a producer.”

That is also why it works the way it does. You bring your own instrumental. You write your own lyrics. The vocal comes out as a stem and goes into your own DAW. And the singer is still in the picture, because in my own work she always was.

Someone else's read on it

Rob Davy is a producer in the UK and a Vokals customer. He wrote this after his first time using it, without being asked for anything:

“Hi Andy, I'm definitely signing up; this is an amazing product. I have no idea how you do it, but I'm certainly curious to know! It is really going to help with getting tracks finished with singers. I believe there is a lot of space to work with AI, and this feels like a very healthy approach to me. Best of luck with it! Best regards, Rob Davy”

I am quoting him because the word “healthy” is his and not mine. Everything above this point is my own argument about my own product. That one sentence is not.

That is where I stand

You can disagree with any of this. I have a stake in the answer and you can weigh what I say accordingly.

What I will not do is pretend the first part of this page is not true.

If you want to see what the tool actually does, it is here: Vokals