Last time, we toured the communities — worship teams, choirs, jazz players, cantors, barbershop singers — and found they had all reached for the same thing: a shared reference to sing against. ( That post is here, if you missed it.) But “backing track” is a deceptively flat phrase. A tanpura humming a single unbroken note and a sixteen-stem studio production of a worship anthem are both, technically, backing tracks. They have almost nothing else in common.
So before you choose one, it helps to know what you’re actually choosing between. There are five questions that, between them, describe any backing track on earth. Ask them in order, and the right source for your music tends to fall out the bottom.
Where did it come from?
Start with origin, because it quietly settles most of the questions that follow. Did you record this yourself at the piano? Buy it from a publisher? Generate it from chord symbols? Pull it from the public domain? Render it from a score into synthesized audio? Each of those is a different kind of thing with a different set of strings attached.
A track you made yourself has the clearest origin — you know exactly where the recording came from, because it came from you. (Whether you are free to publish it is a separate question, and a trickier one, as we are about to see.) A commercial stem came from a company that recorded it on purpose and expects to be paid. A public-domain melody came from a composer who died long enough ago that the law has let go. And an AI-generated bed came from a model trained on who-knows-what. Same four-minute audio file, four completely different origin stories — and origin is what determines the next question.
Can you actually use it?
This is the nervous one, and most people ask it too late — after they’ve already posted the recording. Sources sort into a handful of buckets, and the gaps between them matter more than they look. Some are clean outright: publish and never think about it again. Some are licensed for a specific use — you bought permission, not the song. Some are a genuine gray area that no one can resolve for you. And one bucket looks clean but isn’t, and it’s the one that catches the most people.
Call it clean-PD. Sit down and play a Bob Dylan song and you own your recording of it — but you do not own the song. Put that recording online and you still need the composition cleared, exactly as the coffee shop covering it would. The master is yours; your right to publish rides entirely on whether the song underneath is free. A self-recorded hymn from 1850 is clean-PD and safe. A self-recorded chart-topper is clean-PD and very much not. Same microphone, same performance, opposite outcomes — because clearance was never about the recording. It was about the work.
Which is why the public domain is trickier than it sounds. A hymn can be centuries past copyright while the specific modern arrangement you’re holding is very much under it — a free tune is not a free track. That’s still clean-PD; you just have to be sure the version in your hands is the free one. The genuine gray area is different, and smaller than people fear: an AI-generated accompaniment whose legal status nobody has settled, where no amount of care on your part resolves it. We’re going to spend the entire next post in this territory — the performing-rights organizations, the worship and synagogue licenses, the maddening split between owning a song and owning a recording of it. For now, the only rule that matters: ask this question before you publish, not after.
Whose tradition is it for?
Some sources are universal. A piano accompaniment you record yourself will serve a Methodist hymn, a Hebrew prayer, a folk ballad, or a jazz standard without complaint. But many sources are deeply specific. A worship stem is built for the contemporary praise repertoire. A cantorial recording carries the melodic grammar of a particular service. A raga drone belongs to a tradition five centuries deep. Knowing whose music a source was made for saves you from forcing a tool to do a job it was never shaped for.
What’s the job?
A backing track that is perfect for one purpose can be useless for another. A part-dominant rehearsal recording — your line pushed forward, ugly but accurate — is exactly right for learning the alto part in the car and exactly wrong for a concert stage. A polished commercial stem is the reverse. So ask what this track is for: learning it, worshiping with it, performing it. The honest answer narrows the field fast, because most sources are quietly optimized for one of those jobs and merely tolerable at the others.
How good does it need to sound?
Finally, fidelity — and here’s the counterintuitive part. Richer is not always better. A lush, fully produced stem is wonderful for performance and actively unhelpful for learning, where a dry rendered line or a bare click tells you the pitch and the timing with nothing in the way. Fidelity runs from a full studio production down to a single metronome click, and the right point on that scale depends entirely on the previous question. Match the fidelity to the job, not to your vanity.
Five questions, one filter
Here’s the useful thing about five questions: they compose. “Show me sources that are clean to publish, in the Protestant tradition, for worship, at solo-voice fidelity” is not a thought experiment. It’s a filter, and it returns a short, specific, actionable list.
So we built the filter. Below is every source type we’ve been able to catalog — from the piano you play yourself to a generated tanpura drone — and you can narrow it by any combination of the five questions. Tap the facets that describe your situation and watch the field shrink to what actually fits.
Live accompanist recording
Clean if Public DomainSelf-MadeJewishCatholicProtestantHindu / SanskritSecularLiturgicalTeachingConcert / PerformanceSolo AcousticLive RecordingYou play it, you own the recording — but publishing it cleanly still requires the song itself to be your own or public domain. Quality varies; always available. The fallback that's never out of reach.
Solo reference voice
Clean if Public DomainSelf-MadeJewishCatholicHindu / SanskritLiturgicalTeachingSolo AcousticA cantor, clergy member, or teacher records a lead voice. The track for when there is no instrument at all — chant, cantillation, responsive reading. The recording is yours; clean to publish when the chant or setting is traditional or public domain, not when it's a copyrighted melody.
Self-made DAW multitrack
Clean if Public DomainSelf-MadeProtestantSecularTeachingConcert / PerformanceLiturgicalStudio StemBuilt in GarageBand or Logic. You own the recording and control everything — clean to publish when the underlying song is yours or public domain. Real production effort on you.
Commercial backing tracks & stems
Licensed UseCommercial ProductProtestantSecularLiturgicalConcert / PerformanceStudio StemMultiTracks, Karaoke-Version, Hal Leonard. You're buying a use, not the song. CCLI and print/mechanical licensing diverge sharply here.
Karaoke / instrumental versions
Licensed UseCommercial ProductSecularSecularTeachingStudio StemLive RecordingOff-the-shelf instrumental versions of popular songs. Convenient and everywhere — licensed for a use, not for republishing your finished mix.
Tradition-specific liturgical publisher
Licensed UseCommercial ProductCatholicProtestantJewishLiturgicalStudio StemLive RecordingA tradition's established sacred-music publishers selling scores and recorded accompaniments — GIA and OCP (Catholic), denominational hymnal publishers (Protestant), Transcontinental Music (Reform Jewish). Convenient and tradition-specific; licensing terms vary by publisher.
Public-domain composition
Clean if Public DomainPublic DomainCatholicProtestantJewishHindu / SanskritSecularLiturgicalConcert / PerformanceTeachingSolo AcousticMIDI RenderThe composition is free to use — but your recording or arrangement of it is a new, separate work, and someone else's published arrangement may still be under copyright. Free tune, not free track.
Public-domain recording
Clean to PublishPublic DomainCatholicProtestantSecularConcert / PerformanceTeachingLive RecordingHistorical recordings that have aged into the public domain (in the US, most recordings published before 1925). The recording itself is free — but it is fixed, old, and not made for your tempo or key.
AI-generated accompaniment
Gray AreaAI-GeneratedSecularTeachingConcert / PerformanceStudio StemSuno, Udio, and similar. Instant and cheap; fine for a scratch reference. The copyright status of the output is genuinely unsettled — risky as a published deliverable.
Notation-to-audio render
Clean if Public DomainNotation RenderCatholicProtestantJewishSecularTeachingLiturgicalMIDI RenderClick OnlyMIDI, MusicXML, or LilyPond synthesized to audio. Total control of tempo and key; the timbre is robotic. Clean to publish when the score itself is public domain. The good-enough reference nobody wants to perform to but everybody can rehearse against.
Institution's licensed library
Licensed UseThird-Party LicensedJewishCatholicProtestantLiturgicalStudio StemLive RecordingMaterial a synagogue or church uploads to its own private library, covered by the institution's existing licensing. The Lyrekos public library stays scrupulously public-domain; private libraries operate under their own policies.
Royalty-free / stock music library
Clean to PublishRoyalty-Free LibrarySecularSecularTeachingConcert / PerformanceStudio StemPixabay, Epidemic Sound, Artlist. Broad, permissive reuse under a library license — often free or subscription. Strong for secular and instrumental beds; thin for specific repertoire or liturgical material.
Music-education accompaniment
Licensed UseCommercial ProductSecularTeachingStudio StemPractice accompaniment built for learning. Music Minus One pioneered solo-minus tracks; SmartMusic (MakeMusic) is the modern successor, pairing accompaniment with assessment. Licensed for student practice, not for republishing your mix.
Community part-learning tracks
Licensed UseCommercial ProductSecularTeachingSolo AcousticLive RecordingPart-predominant learning tracks from the Barbershop Harmony Society — each voice part recorded so a singer can learn their line. The closest existing analog to what Lyrekos does, which is worth noticing.
Generated drone & rhythm reference
Clean to PublishSelf-MadeHindu / SanskritLiturgicalTeachingConcert / PerformanceSolo AcousticA continuous tonal drone (tanpura) plus a rhythmic cycle (tabla/lehra), generated in any key and tempo from apps like iTablaPro, iShala, or TaalMala. No song and no score — just a shared tonal and temporal reference the singer locks to. Arguably the oldest expression of the Lyrekos idea.
A few worth knowing by name
The grid will show you the whole landscape, but a handful of sources are worth a word on their own.
The one that’s always available is the one you make yourself — a phone recording of a piano, a guitar, or just a steady lead voice. It will never win an award. And as long as what you’re playing is your own or in the public domain, it will never get you a copyright letter either. Record someone else’s song, though, and that protection evaporates: the recording is yours, but the song is not. For a great many real situations, played from clean material, it is simply the right answer.
The commercial stem is the opposite trade: studio-grade sound, in exchange for licensing terms you have to actually read. The public-domain entry is the trap we keep flagging — the tune may be free while the arrangement is not, which is the whole reason the next post exists. The AI-generated bed is fast, cheap, and legally unsettled, which makes it fine for a private scratch track and risky for anything you publish. And the generated drone — a tanpura and a tabla cycle conjured in any key from an app — is the oldest idea in this entire catalog, a continuous reference with no song and no score, which is about as close to the pure principle as it gets.
There is no best, only fit
Notice what the grid will never do: tell you which source is best. That’s not an oversight. There is no best backing track, only a best fit — the source that’s clean for your use, in your tradition, made for your job, at the fidelity that job needs. Get those four lined up and the fifth, where it came from, has usually answered itself.
And then comes the part we care about most. Once you’ve chosen your reference — whichever of these it turns out to be — Lyrekos only needs one thing: that everyone is locked to the same one. The drone, the stem, the click, the piano you recorded last night — any of them works, as long as your singers are all following it together. The hard problem was never the track. It was getting everyone to hold onto it at the same instant, across a slow and crowded internet. That’s the part we solved.
Next in the series: the question we kept deferring — who actually owns the song you’re singing, and what you’re allowed to do with it. The licensing post is where the gray areas get their reckoning.
Want to put a backing track to work with singers in different places? Give it a try and sing together, wherever your voices are.

Lance Glasser
Lance is CEO and Co-founder of Kinetic Audio Innovations. He was previously a faculty member at MIT, Director of Electronics Technology at DARPA, and CTO at KLA. He also makes sculpture, which has nothing to do with audio but explains the hundreds of pounds of bronze in his house.
Sing in Sync, Anywhere
Whatever reference you choose, Lyrekos keeps every singer locked to the same one — across any distance.
Join our CommunityTry it