How to Make a Remix Using Stems

To make a remix using stems, you split the original song into its separate parts — vocals, drums, bass, and other — with an AI stem splitter, then keep the pieces you want (usually the vocal), replace or add your own, match everything to one key and tempo, and rebuild the arrangement in a multitrack editor. A remix is the same song re-imagined: the recognizable vocal, a new backing. Vocal Cut's browser tools prepare each part — separation, key detection, tempo and pitch work, trimming — on files you already have, and because it all runs locally, your audio never leaves your device. The final layering happens in a DAW. Here is the whole workflow.

1. Split the song into stems

A stem is one isolated layer of a mix — the vocal by itself, the drums by themselves, and so on. A finished record hands you none of these; you have to pull them apart from the stereo file.

Open the stem splitter, drop in the song, and let it separate the track into four stems: vocals, drums, bass, and other (everything else — guitars, synths, keys). It runs one AI model (HTDemucs) that has learned what each instrument family sounds like and reconstructs each one on its own. Once the stems appear you can solo, mute, balance, and download each part individually, so you can pull out exactly the layers your remix needs.

Two honest expectations. First, it's fast, local, and free, but not instant: separation runs on your own device's CPU, and the very first time you use it the model downloads (tens of MB) before processing starts. After that it's cached, and a full song takes real time to process. Second, feed it the cleanest source you can. Stereo files separate better than mono, the current per-file limit is five minutes, and it accepts MP3, WAV, FLAC, and M4A. A lossless WAV or FLAC gives the model far more to work with than a low-bitrate MP3.

If all you actually want is the vocal on top of an instrumental — no drum or bass isolation — the two-stem vocal remover is the quicker route. For a fuller comparison of the two approaches, see stem splitting vs vocal removal, and for the step-by-step on separation itself, how to split a song into stems.

2. Choose your remix direction

Before you rebuild anything, decide what kind of remix you're making — this determines which stems you keep and which you throw away.

Most remixes start from the acapella: you keep the isolated vocal and build a completely new backing under it. That's the flexible, high-impact choice, because the voice is what listeners recognize and everything underneath is yours to design. So the common recipe is: keep the vocal stem, discard the original drums and bass, write your own beat and bassline, and decide whether any of the "other" stem (a signature riff, a synth hook) is worth re-using.

But that's not the only direction. You might keep the original drums and bass and only replace the melodic top. You might isolate a single element — see how to isolate an instrument from a song — and sample just that. The point of separating first is that every layer becomes an independent ingredient you can keep, drop, or rework.

3. Find the key and tempo of your parts

Anything new you add has to agree with the stems you're keeping, and two properties decide that: key and tempo. Measure them before you start writing.

Key is the set of notes the vocal is built around. Run the vocal stem (or the original song) through the key finder so you know what your new chords and bassline have to sit in. Tempo is the speed in BPM — find it with the BPM finder so your new drums lock to the vocal's phrasing rather than drifting against it. Matching key and tempo is the same discipline DJs use; beatmatching and key matching explained covers why both matter together, and how to find the BPM of any song goes deeper on tempo detection.

If you're building the new backing from scratch, this is easy: write it in the vocal's key and at its BPM from the start. The matching problem only gets hard when you're combining the vocal with material recorded at a different key or tempo — which is the next step.

4. Adjust a stem's key or tempo if it won't fit

Sometimes a stem you want to keep is in the wrong key, or a loop you're borrowing runs at the wrong speed. You can move either one, carefully.

To shift key, use the pitch changer; to change speed without touching pitch, use the speed changer. Keeping those two independent matters — an old turntable would speed a track up and raise its key together, but modern time-stretching lets you fix tempo without undoing the key work you just did.

Move by the smallest amount that works. A semitone or two of pitch shift is nearly transparent; push four or five and it starts to sound unnatural. Vocals are the most fragile thing to shift — go too far and they turn chipmunky or hollow. How to pitch shift without ruining vocals covers keeping the shift clean. As a rule, pick source material that's already close to your target rather than relying on a big correction.

5. Trim and loop the parts you're keeping

Raw stems rarely start where you want them. Before they go into the arrangement, clean up their edges.

Use the audio cutter to crop a stem to the exact section you want — the chorus vocal, a four-bar drum fill, a single sung line — and to shave off lead-in silence so the part starts on the downbeat. If you're stitching sections into a longer structure, or looping a phrase back to back, join them end to end with the audio joiner. The goal is that every ingredient is trimmed to length and ready to drop straight onto a track.

6. Build the new arrangement, mix, and export

Layering the prepared stems into a new track, sliding them into place, and bouncing the result to one file is the step the single-purpose browser tools don't do for you. That final assembly happens in a multitrack editor or DAW: put the kept vocal on one track, your new drums and bass on others, and arrange them into an intro, verse, and chorus. Line up the downbeats so the vocal's phrasing lands where you want it against the new groove, then set levels so the vocal sits clearly on top without fighting the beat.

Vocal Cut's job is to get each part ready to drop in. Once you've bounced the arrangement to a single file, you can bring it back to the browser for finishing — audio mastering for glue and competitive loudness, or the volume normalizer to even out the overall level — then export. WAV keeps it lossless for further editing; MP3 gives you a smaller shareable file.

If you want a rundown of tools built specifically for this first separation step, best free stem separation tools compares the options.

A note on copyright

A remix is built from someone else's recording, so it's copyrighted material by definition. Making one for practice, learning, or personal creative use is generally low-risk. Publishing, streaming, or monetizing it is a different matter and usually needs the rights holders' permission — unless the source was released for remixing, as official remix competitions and stem packs are. This isn't legal advice; copyright and fair use when sampling and remixing walks through where the lines tend to fall.

Frequently asked questions

Can I make a remix using stems for free? Yes. Splitting the song into stems, finding key and BPM, shifting pitch and tempo, and trimming the parts are all free in your browser with no account and no software install. Processing runs on your own device, so your audio never leaves it. You then layer the prepared stems into the new arrangement in a multitrack editor for the final mix.

Do I need a DAW to remix with stems? For the final assembly, yes — layering multiple stems to play at once and bouncing them to one file is the job of a multitrack editor or DAW. Vocal Cut handles everything up to that point: separating the stems, matching key and tempo, and trimming each part. A free DAW is enough; you don't need paid software to arrange a handful of prepared stems.

Will the stems sound perfectly clean? Not perfectly. AI separation estimates each part rather than recovering the original studio stems, so dense, heavily processed, or low-quality mixes leave some smearing and bleed between layers. A clean vocal over a sparse arrangement comes out very usable; a busy, effect-drenched mix carries more artifacts. Starting from a lossless, stereo source gives the model the best chance.

Is it legal to remix someone else's song? Making a remix for personal practice or creative learning is generally low-risk, but it uses a copyrighted recording, so publishing, streaming, or monetizing it usually needs the rights holders' permission — unless the material was released for remixing. This isn't legal advice; read the copyright guide for a fuller picture before you distribute anything.