How do I make a karaoke track?

Two routes, and the difference is what you get back. Exclude the vocal returns one file: the whole song with the vocal removed. The Karaoke preset returns 2 files — the instrumental and the vocal separately — which is what you want if you also need the vocal, for a guide track or to check the removal.

How do I make a practice track for my own instrument?

Exclude your instrument, if it is one of the 8 the tab offers: Vocals, Drums, Bass, Guitar, Piano, Other, Saxophone and Wind. You get the full arrangement with your part missing, which is a better practice tool than an isolated stem — you are playing with the band rather than along to a fragment.

Which stems can be converted to MIDI?

Bass, Vocals, Piano and Guitar — the pitched parts. Guitar is marked Experimental, and Drums aren't transcribed yet — coming later. Everything else has no Convert to MIDI action at all rather than an action that disappoints.

How good is the MIDI?

MIDI is a starting point — expect to tidy timing and note lengths. Drums aren't transcribed yet. Dynamics are approximate. Works best on clean recordings. Of the eligible parts, bass is the one to start with — the tool's own note on it is “Transcribes most reliably — the best place to start.” Treat any of them as a head start on transcription, not a finished score.

What does Open in Studio do?

It loads the audio into AudioPod's Studio instead of downloading it. On a single row, function(){throw Error("Attempted to call OPEN_IN_STUDIO_LABEL() from the server but OPEN_IN_STUDIO_LABEL is on the client. It's not possible to invoke a client function from the server, it can only be rendered as a Component or passed to props of a Client Component.")} sends that one part. On the job, function(){throw Error("Attempted to call OPEN_ALL_IN_STUDIO_LABEL() from the server but OPEN_ALL_IN_STUDIO_LABEL is on the client. It's not possible to invoke a client function from the server, it can only be rendered as a Component or passed to props of a Client Component.")} sends every part as its own track, which is the quickest way to balance a split into a rough mix.

Can I use these stems in a release?

Splitting a recording does not change who owns it. Stems from someone else's song carry exactly the rights the song did, and a remix, a cover bed or a sample still needs whatever permission it would have needed before. This is a technical tool, not a licensing one — check the rights for anything you publish.

Stem Splitter

What to do with stems

The four jobs people actually use this for, the fastest route to each, and the two things you can do to a finished part without downloading it — convert it to MIDI, or open it in the Studio.

Lesson 4 · advanced · 7 min read

Open Stem Splitter

Almost everyone arrives at this tool with one of four jobs in mind. Each has a fastest route, and in every case the route is decided by picking the right mode rather than by anything you do afterwards — which is why this is the last lesson rather than the first.

The four jobs

What you want: How to get it
A karaoke or backing track: Exclude the vocal for one ready-mixed file, or the Karaoke preset if you also want the vocal itself.
A practice track for your instrument: Exclude your own part. You get the full arrangement with a hole where you play, which is far more useful than an isolated stem of it.
A remix or a sample bed: Band or Detailed to get the parts, or a decomposition package if you need the kit broken up. Isolate when you only want one element.
A podcast or interview with music under it: Exclude to pull the music bed out and leave the speech. For two people talking over each other, that is a different tool — speaker separation, not stem splitting.

What people use stems for, and the shortest way there.

Tip

Notice how many of these are Exclude. Any job you can phrase as “the song without the ___” belongs there, and phrasing it that way out loud is the fastest way to pick the right tab.

Convert a part to MIDI

Every eligible finished stem carries a Convert to MIDI action underneath its player. It transcribes the notes in that part into a .mid file you can drop into any DAW or notation software — useful when what you needed was not the audio but the notes: a bass line to learn, a melody to re-voice, a chord progression to build on.

Stem: What to expect
Bass: Transcribes most reliably — the best place to start.
Vocals: Lead melody line.
Piano: Chords and melody.
Guitar (Experimental): Dense or heavily-strummed guitar is hit-or-miss.

The parts that offer MIDI conversion, with the tool's own notes.

Drums aren't transcribed yet — coming later. Anything else — the “other” part, ensembles, effects — has no conversion action at all, because a transcription of them would not be worth the credits.

Careful

MIDI is a starting point — expect to tidy timing and note lengths. Drums aren't transcribed yet. Dynamics are approximate. Works best on clean recordings. It is a starting point and the tool says so on the button. Converting is billed separately from the split, at 300 credits per minute of the part you convert, and the estimate appears before you confirm.

Start with bass if you have never done this. Transcribes most reliably — the best place to start. It is monophonic, it is loud in the mix, and its notes are unambiguous — everything that makes transcription reliable.

Send it to the Studio instead

The other per-part action is the Studio handoff, and it is the one that most changes how the tool feels to use. Rather than downloading files and opening them somewhere else, you can load them straight into AudioPod's Studio.

Every stem downloads on its own, from its own row. There is no zip and no download-all.
function(){throw Error("Attempted to call OPEN_ALL_IN_STUDIO_LABEL() from the server but OPEN_ALL_IN_STUDIO_LABEL is on the client. It's not possible to invoke a client function from the server, it can only be rendered as a Component or passed to props of a Client Component.")} loads the whole job into the Studio as one track per stem — which is the fastest route to a rough mix, and it downloads nothing.
function(){throw Error("Attempted to call OPEN_IN_STUDIO_LABEL() from the server but OPEN_IN_STUDIO_LABEL is on the client. It's not possible to invoke a client function from the server, it can only be rendered as a Component or passed to props of a Client Component.")} on a single row does the same for one stem.

function(){throw Error("Attempted to call OPEN_ALL_IN_STUDIO_LABEL() from the server but OPEN_ALL_IN_STUDIO_LABEL is on the client. It's not possible to invoke a client function from the server, it can only be rendered as a Component or passed to props of a Client Component.")} is the version worth knowing. It takes the whole job and lays every part out as its own track, aligned, ready to balance or mute or bounce. For a remix bed or a rough mix it removes the entire download-and-reimport step, and for judging a split it is simply the quickest way to hear the parts against each other.

What stems are not

Two honest limits, both worth knowing before you build a project on this. A stem is not a multitrack: it is an estimate of a part, carrying whatever bleed and artefacts the separation left behind, and it will not survive the same treatment an original recording would. Push one hard with EQ or a heavy effect and the artefacts come forward.

And separating a recording does not change who owns it. Stems taken from a commercial song carry exactly the rights that song carried — a remix, a cover bed or a sample needs the same permission it always did. AudioStems is a technical tool and cannot grant you any right you did not already have.

Questions

What people ask about this

Free to start

Now go make one

Reading about a style description only gets you so far. The studio is free to use — write one sentence and hear what comes back.

Create a free account

1,000 credits every month. No card required.

All Stem Splitter lessons