Before Anyone Else Hears It — Part Two

How Generative Music Could Give Every Member of the Creative Team Room to Be Wrong First

Part Two of a two-part look at the work that happens before a score is written.

I have always thought of scoring a scene as solving a puzzle. There may be several good answers buried inside it, but the work involves finding the one where all of the pieces fit together best. I would watch the scene repeatedly, absorbing the performances, pacing and emotional turns. Then I would step away from the picture and begin writing, usually starting with a sound, a texture or the beginning of a theme—something that inspired me and might open the door to the musical world of the scene.

Once an idea had begun to take shape, I would bring it back to picture and see how the pieces fit. Some parts might feel right while others might need adjusting. I could change the pacing, move the entrance or follow the idea in a different direction until I understood what it wanted to become.

That early stage benefited enormously from privacy. An opinion offered during the inception of an idea can derail it before the person creating it has had time to understand where it might lead. Something unusual may be dismissed too quickly, or a promising idea can become self-conscious before it has had a chance to develop.

In my studio, I could play with the clay without explaining every shape it briefly took. I had room to flop, flounder and fail, circle back to an earlier thought, or keep turning the pieces until the music fit the scene. By the time anyone else heard it, the most fragile part of the process had already taken place in private.

Directors, producers, editors, showrunners and music supervisors have rarely had access to an equivalent space when forming their own ideas about music. They may have strong instincts about a scene, but those instincts do not always arrive fully formed or easily expressed. In a spotting session, they often have to work through those thoughts in real time, in front of everyone else.

An incomplete idea can be misunderstood or redirected by someone who speaks more confidently. Someone without much musical vocabulary may hold back even when their instinct is valuable. They may rely on the temp score or wait to react to somebody else’s suggestion because they have never had a private place to explore their own.

Generative music may finally provide that space.

That possibility is what led me to build CuePilot™, a scene-oriented prototype designed to give each member of the creative team room to work through a pivotal scene before bringing their ideas into the larger conversation. At this stage, I use it as a single-user tool. A person can enter and adjust information about character motivation, dramatic intention, point of view, timing and relevant pages from the script. CuePilot then organizes those choices into prompts used to generate several musical approaches to the scene.

I designed this stage of the workflow to happen before the group attempts to reach a collective decision. A producer, editor or director can adjust the inputs, change the prompt and hear several approaches against picture. One may clarify an instinct they already had; another may reveal that the emotion is right but the pacing is not. Working alone gives each person time to follow an idea, question it and decide which parts are worth bringing into the room before outside reactions begin shaping it.

Imagine a pivotal scene in which a woman comes home after learning that her marriage is over and begins making dinner as though nothing has happened. The editor may feel that the music should wait until she is alone at the sink. The producer may want the audience to feel the collapse before she allows herself to feel it. The director may hear less sadness than determination. Each could use CuePilot to test that instinct through several musical approaches and bring one or two useful examples into the room. Heard against picture, those examples might reveal that everyone agrees on the late entrance but not on point of view, or that the restraint of one approach and the pacing of another belong in the same final direction.

No one is going to do this for every scene. A film or television episode contains many scenes, and everyone involved already has their time spread thin. The investment would make the most sense with moments that are likely to establish the larger musical direction, for pivotal sequences, or for scenes with unresolved questions. An hour spent exploring one of those scenes may prevent days of writing and revision around an idea the group never fully worked through or understood.

After that independent work, the spotting process could begin with a listening session. Any member of the creative team could bring in one or two selected approaches, allowing the group to hear how several people separately understood the same scene before discussing it together. Once music generators can reliably follow CuePilot’s designated start points, durations and structural instructions, the editor should be able to simply drop each cue directly into the timeline at its precise sync point without first cutting it to fit like a conventional temp track.

The editor could also play the cues without identifying who created each one. Initial reactions would be less affected by title, authority or ownership, and the identities could be revealed afterward when the thinking behind each approach becomes useful.

Several people may arrive with remarkably similar pieces despite having described the scene differently. Their cues could reveal a shared pace, emotional distance or point of view that nobody had been able to articulate in words. Widely different pieces would expose the opposite: the members of the team may be experiencing the scene in fundamentally different ways. Either result would give the group useful information before the composer begins writing.

The individual examples would not need to work from beginning to end, because the purpose would not be to choose a winning cue. One version might nail the emotional register of a scene while dragging through its middle. Another might handle a difficult transition beautifully but miss the tone entirely. A third might capture something about a character’s state of mind that the others overlooked. By identifying those partial successes, the group could begin to describe a direction that no single example had solved completely.

That would make the spotting conversation considerably more specific. The group could discuss where the music should begin, when it should pull back and how its movement should interact with the edit. Everyone would be responding to choices they had just heard against picture rather than trying to interpret broad descriptions that may mean something different to each person.

The composer would then synthesize those observations into one coherent musical idea. The generated examples may reveal the dramatic behavior people are responding to, but the larger musical language and the relationship between those elements still have to be developed. Turning those discoveries into a cohesive score would remain the composer’s work.

This process may also have an unexpected effect on one of the longest-standing problems in film and television music: Temp Love. Temp Love develops when a temporary piece of music is attached to a scene and heard repeatedly until its timing, instrumentation and emotional gestures begin to feel inseparable from the picture. By the time the composer delivers an original cue, anything different can feel wrong simply because it is…different.

The problem is not always that the temp is great—there may only be one or two elements that are working, and some that are not. Yet often, everyone has become accustomed to hearing the scene only one way. Hearing several genuinely different approaches before the score is written may loosen that attachment.

The group might discover that its attachment to the temp has less to do with the cue as a whole than with the one or two specific functions it performs. Perhaps it enters at the exact moment the scene turns, leaves the right amount of space around a performance or gives the edit its pace. Hearing other cues perform those functions differently could help the group preserve what matters without preserving the temp itself. The process would still require discipline: the examples should be limited, meaningfully different and discussed according to what each reveals about the scene.

For most of my career, I had the privacy to search for the score before anyone could react to an unfinished idea. CuePilot extends some version of that protection to everyone responsible for shaping the music. By the time the room comes together, the people involved may have already tested their instincts, discovered where they align and learned which parts of several imperfect cues actually matter.

Together, they will have created the pieces of the puzzle; the composer will still have to determine how those pieces fit together—and, in doing so, solve the score.

© 2026 Rick Butler. All rights reserved.