Blog
AI in the marching arts: more hands, not fewer people
AI has a bad name in this activity. Here is exactly what we use it for, what stays with people, and why the doors are shut while we build.
By The Countoff team · · 6 minute read
AI has a bad name in the marching arts, and it has earned some of it. People in this activity have seen generated music, generated artwork and big promises about software that designs the show for you. A director who has spent twenty years learning how to stage a ballad does not want to hear that a program will do it, and neither does the arranger who wrote the ballad.
We are building Countoff, and Countoff uses AI. So we owe you a plain answer about what that means. Here it is.
People make the show
On Countoff the music is composed and arranged by people. The drill is designed by people, and the dots are placed by the designer, with help from the software when they ask for it and never without them. Students are taught by people. Performances are judged by people, and the score is the judge’s.
None of that is a compromise we made on the way to something else. It is the point. The reason this activity is worth a summer of long days is that people make it, for an audience of people, and are judged by people who did it themselves.
What the software is for
What wears a staff down is rarely the art. It is everything around the art: the tapes nobody has time to listen to, the coordinate sheet that has to be retyped, the same note written out six times for six sections, the step size somebody forgot to check before the set went to the field.
That is where we use AI: as more hands. A third, fourth and fifth arm for the people who already know what they are doing.
- A judge talks through your show. The tape is typed up, with the words this activity uses spelled the way we spell them, and each comment is placed on the set it is about. You read it, search it and decide what matters.
- You photograph a printed coordinate sheet. The dots are read off the page and shown to you to check before they go anywhere near your show.
- A director takes a tape meant for staff and turns it into a short, plain note for the trumpets or the guard. It goes out when the director sends it, and not before.
- A designer says “move the trumpets to the 50 in a circle.” The studio shows what that would look like. Nothing on the field changes until the designer applies it, and one undo takes it back.
- The studio points out a step that is too big for the counts, or two performers who would walk through each other. The designer decides what to do about it.
In every one of these, a person asked, a person checks, and a person decides.
This activity has done this before
Drill used to be written on graph paper. Then it was written in software. Nobody watching a show today credits the software with the design. Designers kept designing, and the drill got harder.
The activity is full of tools like that, and none of them does the skilled part:
- A tuner tells you that you are flat. You still fix it with your ears and your air.
- A metronome keeps time for a whole field. Nobody thinks it performed.
- A tower gives the same pair of eyes a better view of the form.
- A speaker on a cart lets one voice reach the back hash.
- Notation software replaced copying parts by hand. Arrangers still arrange.
- A powered screwdriver and a manual one put in the same screw. The hand and the judgment about where it goes are the same. What changes is how much of your evening is left.
That is the category we want Countoff to be in. A very good multi-tool in the hands of someone who knows the job.
The judge tape is the clearest example
A contest hands a staff several tapes, each one as long as the show. They arrive late on a Saturday. The honest truth in a lot of programs is that some of them are heard once, on the bus, by one person, and some are not heard at all before the next rehearsal.
Nothing about that is a judging problem. The judges did their job. It is a paperwork problem, and paperwork is what software is for.
Today, a tape recorded on Countoff is transcribed, lined up with your drill by count, and summarized by section, so the comment about the transition into set 42 is sitting on set 42. Staff can search a whole season of tapes and ask what was said about the guard in the ballad.
Here is what we are building next, and want your help getting right. When a judge mentions the guard, or a dropped rifle, the person who runs the guard gets that comment, as something to review, without waiting for somebody else to forward it. The same for the battery, the front ensemble and the brass. And when the tape is ready, the right people are told it is ready. Our goal is that this happens within a few minutes of the performance, while the band is still coming off the field.
That part is not built yet, and we have not measured how fast we can make it. We are saying it out loud because it is the kind of thing we think AI is for in this activity: the judge still judges, the staff still decide what to rehearse, and the wait between the two gets shorter.
What we will not do
- Nothing changes in your show unless a person applies it. Suggestions are shown first and can be undone.
- AI does not score a performance. Summaries are a starting point for staff. They can be wrong, and they never stand in for the judge’s sheet or a director’s judgment.
- We do not use student information or judge recordings to train AI models, and the AI services we use are not allowed to train on what we send them.
- You do not need the AI to use Countoff. Writing drill, handing every performer their dot and their music, and running a rehearsal all work without it.
- We say where AI is used. A summary written by it is labeled as one, so nobody mistakes it for the judge’s own words.
Why the doors are shut
Countoff is in closed beta. You cannot sign up; you ask for access, and we bring programs in a few at a time.
That is on purpose. We do not want to push a product at this activity that the people in it have not tested and argued with. Directors, designers, arrangers, techs, judges and the students holding the phone should be the ones who tell us what is useful, what is annoying and what crosses a line. We would rather hear “do not ever do that” from a few programs now than find out from every program later.
If you have an opinion about any of this, we want it, including the opinion that we have something wrong. Ask for access and tell us what you would use, what you would never use, and what you are tired of doing by hand.