In a town this size there is no sound department. There is a person. Around here that person books the band, hauls the speakers, runs the board, and on Sunday afternoon puts together the music for somebody’s memorial slideshow.
I have been that person for about nine years. What follows is one ordinary weekend, and the three or four minutes of laptop work behind each part of it. A vocal remover does one of the three jobs below; the other two need different tools entirely.
Thursday: the karaoke list nobody has
The bar runs karaoke twice a month. The commercial track service covers maybe two thousand songs, which sounds like plenty until three people in a row request something local, something new, or something country from 2019 that never got a karaoke version.
The old answer was “sorry, not in the book.” The current answer takes about ninety seconds. Hand the song to an online vocal remover, keep the side without the singer, drop it in the queue. Nothing to install, nothing to buy, and it all runs in a browser tab on the same laptop I use for the slides.
Two habits I picked up the hard way. Audition it loud on the actual PA, never on laptop speakers, because the room is what exposes a weak separation. And never judge one from the opening bars — songs start sparse, any vocal remover sails through that stretch, and the chorus is where trouble shows up.
Songs built on stacked harmony fight it. Give a vocal remover a heavily reverbed vocal and the backing loses some of its air along with it, because the tail of the voice is smeared across the whole mix. Plain arrangements come back clean enough that nobody at the bar has ever once remarked on it.
Friday: the bass player cannot make it
Wedding gig, four-piece, and the bass player’s kid has a fever. A vocal remover is no help at all here, which is worth saying plainly, because people assume one tool covers every emergency.
Our keys player can cover a bass part with her left hand. What she cannot do is lift it off a record in the ninety minutes between the phone call and load-in. Working a part out by listening takes an afternoon she did not have. She needed those notes on a screen.
So I ran the track through a converter that turns a recording into MIDI notes and set the output to melody rather than polyphonic, which is the right choice when you want one line and not a wall of chords. What comes back is a piano roll — pitch running up the side, time running across — and thirty seconds of squinting tells you whether the detection heard the same song you did. When it hears too much, there is a cleanup setting that thins the result out. I used it twice that afternoon.
Some honesty about that. It reads what is loudest and clearest, so a bass buried under a kick drum comes back with guesses in it, and the fix is a human looking at the grid and deleting the notes that were never played. It took her about ten minutes, which was ten minutes she had. Slides and bends do survive the conversion, which surprised me the first time — a note that scoops into pitch stays scooped rather than snapping to a grid square.
Nobody at the reception knew. The keys player got a lead sheet out of it that she still has.
Saturday: the slideshow problem
Here is the job that used to swallow an entire evening.
Family wants a memorial slideshow. They want music under it. What they hand me is a phone video with a song playing in the background, or a request for a specific recording that the funeral home has no license to play through their system, or nothing at all and a shrug.
For a while my answer was the same four royalty-free piano pieces on rotation, which after enough services starts to feel like I am not doing the job. Now I generate original music by writing down what the piece is for and what it should feel like. Each run hands back a pair of takes rather than one, and that turns out to matter: I lay both under the pictures and let the family react, instead of deciding for them.
There is a toggle for instrumental only. Use it — the same instinct that makes a vocal remover useful on Thursday applies here in reverse. A generated voice singing over a slideshow where a family is trying to hear their own thoughts is exactly wrong, and I learned that the way you would expect.
What I will not do
Two lines I hold, and small-town word travels fast enough that holding them matters.
Pulling a singer out with a vocal remover buys me no rights whatsoever. Karaoke in a licensed bar is covered by the bar’s license, and that is the bar’s business to keep current. A track I built for the karaoke queue does not go home with anyone, does not get posted, and does not turn into somebody’s YouTube upload. That distinction is the whole thing.
I tell people when a file is a workaround. The bride on Friday knew the bass part came off a record. The family on Saturday knew the music was generated rather than licensed from a composer. Nobody has ever objected. What people mind is finding out afterward.
The part that actually changed
Nine years ago the constraint on all three of those jobs was studio access. If you needed a version of something that did not exist, you either knew somebody with a rig or you did without, and mostly around here you did without.
That constraint is gone, and it went quietly. No announcement, no new gear in the truck. Just a browser tab open on the same laptop that runs the slides, and a Thursday where the guy who wanted a song from 2019 got to sing it.
Which is a small thing. But the vocal remover tab, the note converter and the rest of it add up across four gigs a weekend, most weekends, and a lot of small things is what the job has always been.