If you take issue with any use of AI, you may not want to read further. While I have scored 35 films with absolutely no use of AI, I have used AI solely to add musicians and singers to totally human written melody and chords and lyrics.
When I score cues for filmmakers/directors, etc, I send them back not just audio files of the cues but also a video file/mp4 of the audio file mixed with the video/scene it was intended for. My intent is not to replace the video editing and mixing process but merely to give them the idea of what it could look and sound like. It provides an avenue for discussion - what they like, what they don't like, what if anything needs to be added, deleted, or changed. I view the video file purely as a collaborative tool - to show them 'what could be'.
So last year I started co-writing music/songs with a wonderful lyricist, Adryan Russ. So far we have written five songs together. Last week we started working on a theatrical song. When I write a song, I start by creating a lead sheet (with Sibelius) of the melody, chords, and lyrics. Then I send the created PDF (of the lead sheet) to the lyricist normally with just the first verse to see if she likes where I am taking it. I want to get the go ahead from her before I continue working on the song. Now one can say that the lead sheet is sufficient. However, what I have discovered is that our collaboration has been enhanced by going one step further of 'what could be' and creating an audio arrangement with vocalists and musicians assisted by AI. Now there is no intention to market it and in fact, in this scenario it hasn't even been finished yet.
Now when the song is done and the lead sheet is completed, the question becomes what to do next. One path is to use it (or a chord sheet with just lyrics) to get it performed by a musician/singer or band. This might be all they need. But in many cases, the promotion/marketing of the song could enhanced by creating an arrangement of 'what it could be'. Again, the tools especially if AI is used is not used with the intent of commercialization, but as an envisioning process.
So our 5th song was a Gospel piece. The question became after the initial lead sheet, what to do next. Orchestral arrangements can be quite time consuming. So too are Gospel SATB vocal scores along with a piano/rhythm section arrangement. And some church/choirs want to do their own arrangements while others want to get completed arrangements with parts to hand out. So consideration must be given on how much human time be given towards moving ahead with a specific arrangement.
So I decided to stop after the lead sheet creation and promote the music with an example of 'what could be'. Along with the lead sheet or as I went one step further with a piano only arrangement, I created an audio file arranged and played by AI created musicians and vocalists. All of these items were part of a pitch and specifically intended for that purpose. If I went any further such as an SATB arrangement, it would have to be done similar to the cue process above - I would create part of a specific SATB arrangement tailored specifically for those particular singers. It would take their singing ranges and abilities into account. For example, it might have to transpose various sections, decide where and when to modulate and even in the case of the pitch I made on Wed leave out a section making it SAT.
In the aforementioned scenario, I probably would not write the whole customization at once any more than I don't deliver a whole score at once (though others I know do). At each major step or milestone, I would create a audio file to enable my stakeholders to envision approximately what a live performance would sound like.
So here is my current 'take'. Since last summer I have been creating audio/music 'visions' of what could be. There is no current 'intent' to commercialize those audio files. Their main purpose is to aid in the promotion of the core melodies, chords, and lyrics.
So what do I call them. This is where things get 'sticky'. I want to stay on the non-political end if at all possible. I used to call them 'demos'. But when people hear that word, they envision human vocalists, musicians, studios, record companies/publishers, etc. Not AI assisted music.
I have decided to refer to them the same way I refer to the musical scores under development that I deliver to my filmmakers/stakeholders - I will call them a "Mockup".
And if you want an example of my Gospel 'mockup', checkout https://soundcloud.com/joelirwin/together-we-will-rise
Nice, which AI programs do you use?
1 person likes this
I respect your perspective, but I see it quite differently.
To me, one of the main reasons a director hires a composer is not simply to produce music, but to bring a unique artistic interpretation. A d...
Expand commentI respect your perspective, but I see it quite differently.
To me, one of the main reasons a director hires a composer is not simply to produce music, but to bring a unique artistic interpretation. A director can already provide references, temp tracks, or describe the emotion they're after. My job is to transform those ideas into something that is uniquely mine.
That's why AI-generated musical mockups don't sit comfortably with me. If the "what could be" is generated by AI, then in many ways the director could generate that concept themselves and simply ask me to imitate it. At that point, I feel the composer's creative voice starts to disappear.
I also think there's a misunderstanding about the word mockup. In film scoring, a mockup isn't an AI-generated vision of a piece—it's the composer's own performance created inside a DAW using virtual instruments. The goal is to make it sound as convincing as possible because, in reality, many productions will never have the budget to record a live orchestra. For many composers, the mockup is the final product that ends up in the film.
To me, developing the skill to create expressive, realistic performances with samples is part of the craft. If AI replaces that process, we risk losing an important artistic and technical skill.
There's also the ethical side. Current generative AI models have been trained on the work of countless composers and artists, many without explicit permission. That makes me uncomfortable supporting a technology that could eventually be used to imitate or even replace the very people whose work trained it.
I'm not against AI across the board. I think it has value for research, organization, technical tasks, or even assisting with production. But when it comes to creating musical expression—the melodies, performances, phrasing, dynamics, and emotional interpretation—that's where I believe the composer should remain firmly in control.
That's just where I draw the line.
João Moreira I completely agree!!!