RECORD IT AT HOME

This is the part almost everyone gets backwards. People spend three hundred dollars on a condenser microphone and record in an empty room with hard walls, and the result is worse than a phone recorded in a wardrobe. The microphone captures whatever arrives at it, and in a hard empty room what arrives is your voice plus a hundred slightly delayed copies of your voice bouncing off the walls, the window, the floor and the ceiling.

Those reflections are the single thing that cannot be undone later. Compression, EQ, tuning, de-essing, saturation — all of that is fixable, reversible, adjustable. Room reflections are baked into the recording at the moment of capture and there is no process that removes them cleanly. A reverb you did not ask for is permanent. That is why the room comes first.

What you are looking for is a small space full of soft things. Soft absorbs. Hard reflects. A closet packed with hanging clothes is genuinely one of the best vocal booths most people have access to, and it costs nothing. Clothes on hangers, at different depths, at head height, are close to an ideal absorber.

Stand in the room and clap once, hard. Listen to what happens after the clap. If you hear a short bright ring, a slap, or the sound hanging in the air for even a fraction of a second, the room is too live. Go somewhere smaller and softer and clap again. You are looking for a clap that just stops — a dead, slightly boxy sound with no tail.

Rank the rooms in most homes roughly like this. A wardrobe or walk-in closet with clothes in it is best. A small bedroom with a made bed, curtains and a rug is next. A living room with a sofa and carpet is workable. A kitchen, a bathroom, a hallway, a stairwell, or any room with bare walls and a hard floor will fight you the whole way, and bathrooms are the worst of all despite how good singing in one feels.

If the only room available is a live one, build a small dead space inside it. Hang a duvet or a thick blanket over a door, stand facing it about a foot away, and put another behind you if you can. You are not trying to soundproof anything — you are trying to stop your own voice coming back at the microphone.

Almost every phone records at a low bitrate by default because the default is designed for voice notes, not music. Changing one setting is the difference between something that survives a mix and something that falls apart the moment it is processed.

On iPhone: Settings, then Voice Memos, then Audio Quality, and set it to Lossless. On newer iPhones this setting sits under Apps rather than in the main Settings list. Do this before you record, not after — it does not retroactively improve anything you have already captured.

On Android: open your recorder app's own settings and look for format or quality. Choose WAV if it is offered. If only compressed formats are available, pick the highest bitrate on the list. Some manufacturer recorder apps cap quality regardless of the setting; a free recorder that offers WAV is worth installing if yours does.

Airplane mode, every time. A call, a message, or a notification arriving mid-take does not just interrupt you — on many phones it stops the recording or ducks the audio for a second. Losing the take you finally nailed to an incoming call is a specific and avoidable kind of pain.

Clear some storage before you start. Lossless audio is large, and a recorder that stops because the disk filled up will not tell you politely.

On a phone, the microphone is usually at the bottom edge, near the charging port. Find out which one your phone uses before you commit to a position — record ten seconds talking at each end and listen back.

Distance is roughly a fist and a half from your mouth. Closer than a fist and the low end becomes unnaturally heavy and every consonant turns into an event. Further than about a foot and the room starts to arrive with your voice, which puts you back in the problem from section one. Somewhere around six to eight inches is the working range for most voices.

Do not sing straight into the mic. Aim slightly past it, so the column of air from your mouth misses the capsule. Plosives — the burst of air on P, B and hard T sounds — are what create those thumps that ruin an otherwise good take, and they are almost entirely a function of air hitting the microphone directly. Singing just off-axis removes most of them for free.

A pop filter helps, and you can make a serviceable one. A single sock stretched over the bottom of the phone, or a piece of tights over a wire coat hanger held a few inches in front, does most of what a bought one does. Both are ugly and both work.

Hold the phone still, or better, do not hold it. Handling noise travels straight into the body of the phone and up into the recording as low rumble. Prop it on a stack of books, tape it to a mic stand, wedge it in a cushion. Anything that keeps your hands off it.

Record acapella. No beat playing anywhere — not in the room, not in your ear. A voice on its own in a very quiet room is the cleanest thing you can send, and it gives whoever mixes it the most to work with. If you can hold the song without the beat, that is what we would rather have.

The reason is bleed, and bleed has no fix. If the beat is playing out loud in the room while you record, the microphone captures the beat as well as your voice. That bleed cannot be separated out afterwards. There is no plugin, no AI tool and no engineer who can remove a beat from a vocal recording cleanly, because the two sounds are added together into a single waveform. Once mixed at the point of capture, they are one thing.

The practical consequence is worse than it sounds. When the finished record is assembled, the beat plays underneath your vocal — so a vocal with bleed puts a second, slightly out of time, much worse sounding copy of the beat under the first one. It phases, it smears, and it makes a professional master impossible.

If you genuinely need the beat to keep time, the fallback is one earbud in, one ear open. Wired is better than wireless because Bluetooth introduces latency that makes it harder to sit in time, though modern low-latency codecs have narrowed that gap. Keep the level in the earbud low — loud enough to sing to, quiet enough that it does not spill out of the bud and into the microphone. If you can hear the beat leaking when you hold the earbud at arm's length, it is too loud.

One ear open matters as well. It lets you hear your actual voice in the room rather than only the version in your head, and singers pitch better with one ear uncovered. There is a reason you see it in every studio photograph ever taken.

Digital audio has a hard ceiling. Below it, everything is recoverable; above it, the waveform is simply cut off and the information is gone. Quiet is fixable. Clipped is not.

Perform at full commitment and watch the meter. You want the peaks somewhere around two thirds up the meter, with real headroom left above them. In numbers, peaks landing between −12 and −6 dBFS is a comfortable target. If your recorder shows no meter at all, record the loudest line, listen back on headphones, and check whether the peaks sound crunchy or distorted. If they do, move back an inch or two and go again.

Resist the urge to record as loud as possible. A vocal peaking at −10 sounds identical to one peaking at −2 after it is mixed, because the first thing that happens in the mix is a level adjustment. The only difference is that the quieter one still has its transients intact.

If your song has a whispered verse and a screamed hook, do not try to solve it with one distance. Record the quiet parts closer and the loud parts further back. That is what a real vocalist does on a real session, and it is free.

This is the part that surprises people. You are not being asked to learn the record you bought and deliver a finished performance of it. You are being asked for ten minutes of your voice, performing, recorded cleanly. What you perform is up to you.

Rap or sing your own material. Perform a song that is already out. Do the same track three times over. Mix several different ones together. Talk for a stretch in between if you want to. None of it is wasted. Ten minutes is the floor rather than the target — if you have twenty, send twenty.

What we do want is you actually performing rather than only speaking. Rapping, singing, or melodic rap if that is where you sit between the two. A voice in performance does things a voice in conversation never does — it holds notes, it pushes, it sits behind and ahead of a beat, it carries the accent and the phrasing that make you sound like you. That is the material that matters.

So do not agonise over song choice, and do not stop and restart every time you fluff a word. Sing or rap through it. Energy is the thing that cannot be added afterwards; a wrong word is nothing. And do not delete a pass because you did not like it at the time — what feels like a bad take while you are in it is regularly the most convincing thing on the file.

Dry means no processing of any kind. No autotune. No reverb. No delay. No EQ. No compression. No noise reduction. No filters, no presets, no enhancement, no vocal effect the recorder app offers you.

Every one of those is a decision baked permanently into the file, and every one of them removes a choice from whoever mixes it. Autotune applied at the source cannot be dialled back if it is too hard, and it cannot be dialled up if it is too soft. Reverb printed onto a vocal cannot be removed to put a different one on. Noise reduction is the worst of them: it works by aggressively removing frequency content, and what it takes with the noise is the air and detail at the top of your voice that makes it sound like a person.

A dry vocal sounds bad on its own. It is supposed to. It sounds thin and close and exposed and slightly awkward, because that is what an unprocessed human voice sounds like. Do not judge your take by how it sounds dry — judge it on the performance, and let the tuning and the mix happen where they can still be changed.

One caveat worth stating honestly: the quality of what comes back depends on the quality of what goes in. A well-performed take in a soft room can be made to sound like a record. A take with a loud fridge, a clipped chorus and beat bleed cannot, by anyone, at any price.

• Fridge, air conditioning, fan and extractor off. Go and physically check.
• Windows shut. Traffic and birdsong both end up in the take.
• Phone in airplane mode, recorder set to its highest quality.
• Storage cleared.
• No beat playing. If you need one to keep time, one earbud in, low enough that it does not leak.
• Phone propped, not held. Roughly a fist and a half away, slightly off-axis.
• Record ten seconds of silence first — this gives whoever mixes it a sample of your room, which genuinely helps.
• Record thirty seconds at full energy and listen back before committing to the session.
• Water nearby. Not dairy, not cold.

• Beat playing out loud. Unfixable. Covered above, and worth repeating because it is the most common single problem.
• Recording in the bathroom. It sounds enormous while you are in there, which is exactly the problem — that is the room, not you, and it is printed into the file.
• Sending two minutes instead of ten. Less material means fewer options, and it shows in the result.
• Autotune applied at the source. Removes every subsequent choice.
• Recording too hot. Clipped peaks cannot be restored.
• Holding the phone. Handling noise as low rumble under everything.
• Standing too close. Boomy low end and plosives on every P.
• Sending compressed voice-note files. Default quality is built for speech, not for a record.
• Deleting the takes that felt bad. Those are frequently the ones that get used.
• Only talking. We need you performing — rapping, singing, or melodic rap. Talking alone does not carry the phrasing.

For a MATRIX MUSIC reference song, the process from here is fixed. Your takes are comped — the best moments assembled into one continuous performance. That comp is tuned, which on a good take is a light touch rather than a rescue. It is then edited for timing, de-essed, compressed, EQ'd, given its effects, and sat into the instrumental. The guide vocal that was on the demo comes out entirely; yours takes its place. The result is mixed and mastered to release standard.

That comes back in one to five business days, and within seven in a busy week. You own the recording. The record comes off the site permanently and nobody else can ever release it — every reference song is sold a single time. The songwriting stays with MATRIX MUSIC, because MATRIX MUSIC wrote it; the full terms are in the licence agreement, printed before you pay rather than buried in a PDF afterwards.

When you are ready, the upload page is at matrixmusic.net/send-vocals. If you have not picked a record yet, the catalogue is at matrixmusic.net/reference-songs — the voice you hear on every demo is a guide, and it is not the one that comes with the song.

• Can I use my AirPods to record? You can, and it will be usable in a pinch, but the microphone in a wireless earbud is optimised for speech intelligibility on a call and it aggressively processes what it hears. The phone’s own built-in microphone in a soft room beats AirPods in the same room almost every time.
• Is a $100 USB microphone better than my phone? In a treated room, clearly yes. In an untreated room, often no — a sensitive microphone picks up more of a bad room, not less. Fix the room first; the microphone is the second purchase, not the first.
• How do I know if my room is too loud? Record sixty seconds of silence and listen back on headphones at a high volume. Whatever you hear — traffic, a fridge, a fan, a neighbour, hiss — is in every take you record there.
• Should I stand or sit? Stand if the song has any energy in it. Sitting compresses the diaphragm and it is audible.
• What about at night? Late is often the quietest the room will ever be, which helps. Voices are also generally lower and less flexible late at night, which does not. Warm up properly and be honest about whether the take has the energy the song needs.
• How long should the whole session take? Ten to fifteen minutes of recording gets you the ten minutes of material. Because you are performing things you already know, there is nothing to learn first.
• Do I need to record to a click? No. Acapella is preferred — perform it the way you feel it and the timing is edited in house. If you are using the one-earbud fallback, the beat is the click.
• What file should I send? Whatever your recorder produced at its highest setting, unedited. WAV or lossless if offered, otherwise the highest-bitrate file available. Do not convert it, do not compress it further, and do not trim it — send the whole thing.
• Can you fix a take with the beat bleeding into it? No. Nobody can. It is the one problem on this page with no solution, which is why it has its own section.