The John Lennon Remixes

First, a short editorial …

Come June of this year (2025), I will have had a 40-year career as a software engineer, a career which has spanned the advent of major computing and software, and therefore human experience, game-changers. I am referring to things like the adoption of personal computers, the Internet, the World Wide Web, smart phones, and social media, not to mention the impact computing and software has had on everything from manufacturing to medicine to finance to travel to, well, you name it. So I feel like maybe I have the experience, expertise, and insight to confirm what many of you have probably already assumed: We are at the precipice of something huge — and I mean ridiculously huge, unlike anything that’s come before — in terms of Artificial Intelligence.

I don’t want to frighten anyone here. I can’t say that I, personally, am frightened by the potential of AI. But I am concerned about how it can be (and in some cases already has been) used to trick people into believing something that is simply untrue.

Could AI be used, for example, to persuade the masses that an attack is imminent and therefore an invasion of a sovereign state is justified, or that (extreme case) a nuclear launch is necessary? Maybe. I still have faith (for now) in humanity’s ability to reason and prevent global catastrophe. Could robots become self-aware and rise up against us, Terminator-style? Again, maybe, but that seems a little far-fetched to me.

I’m more concerned about “deep fakes” that could result in the election of corrupt and dangerous political leaders — at all levels — that threaten our democracy, our freedoms, and the way of life we’ve come to take for granted.

History is full of never-realized doom-and-gloom predictions saddled to technological advancements, so I’m mostly going to side with history on this one. It is 100% true that AI has the potential to bring enormous amounts of “good” into people’s everyday lives, on a scale we’ve never experienced. But it will still require humanity’s ability to use logic and analysis and contemplation and reasoning to control it.

Now, on to something related, but much (MUCH!) more frivolous …

Short version: I took 2 of my original recordings, and I remixed them using a vocal track provided by a John Lennon AI voice model. Please refer to the “deep fake” point I made above. Yikes!

There are millions of AI voice models out there on the Interwebs. As is always the case, some are better than others. Many are models of famous people (and not always singers), but most are not. If, for example, you wished to generate an audio file of SpongeBob SquarePants reciting your dissertation, you could do so very easily. On the other hand, if you are a musician and songwriter and not a particularly great singer, you can find some very nice, non-famous voice models out there to give your vocal track exactly what you feel is needed to realize your vision.

I did not spend a lot of time comparing the various John Lennon models. I wanted to do something quick and simple-ish, so I found a service that had a decent-sounding option that also met my needs, and I just went with it. If I had invested more time and energy, perhaps my results would have been better. Not sure.

There are super easy ways to do what I set out to do, and had I gone that super easy way, the results probably would have been crappy. I elected to put a bit more effort into this, and my results are very much less crappy, though not even close to outstanding. “Where Did You Come From?”, which kind of had a Lennon-vibe to begin with, came out pretty good. “Give Me Back My Dreams” didn’t fare as well, but it seemed to improve and sound a bit more like Lennon by the second verse.

The first step in the process was to produce audio files that had nothing but the isolated lead vocals for the original tunes I chose. The next step was to use an online service to create AI-generated John Lennon voice model versions of those vocal tracks. The final steps were to remix and remaster my songs, incorporating those AI-generated vocals.

In the end, the AI models did introduce some artifacts that future (or different) models won’t (or don’t). For example, there’s a tininess to the vocals that I had to work on to lessen. There are also some unexpected momentary increases and/or decreases in volume that I mostly just ignored. In “Where Did You Come From?”, there’s a lyric “but the love I feel” that seemingly got transposed to “but the love I deal.” That sort of thing. Still, I felt the results were worth sharing, so here they are.

For reference, here are the original releases of Where Did You Come From? and Give Me Back My Dreams. And here are the John Lennon remixes:

I think “Give Me Back My Dreams” might be more interesting with an AI-generated Bono or Elvis Costello vocal, neither of which I found on the service I used, but their library is always expanding. Stay tuned!