All true, and none of it is the point. A site is not a discovery channel and it never was. It is the one address in the whole operation that cannot be taken away, rate-limited, deprioritised, or quietly redesigned around you.

What you actually own

Count what an independent artist controls outright. A domain. A mailing list. The files themselves. That is close to the whole list. Everything else — followers, playlist placements, monthly listeners, video views — is held on your behalf by a company whose product roadmap you do not see and whose terms you agreed to without reading.

None of that is a conspiracy. It is just the arrangement. Platforms are extraordinary distribution and terrible property. The lesson every artist learns eventually is to use them hard for what they are good at, and to keep the things that matter somewhere else.

What a site is for

Four jobs, in rough order of importance:

• Being the answer to “where do I find your stuff?” One link that leads to everything, rather than five that each lead to a fifth of it.
• Collecting emails. The only audience list you can export, back up, and take with you.
• Being the place press, promoters and collaborators land. A booking agent looking you up wants a bio, photos and contact details in ten seconds, not a scroll through your grid.
• Saying things in your own words. A caption is not writing. A site is where the longer version lives.

What actually belongs on it

Most artist sites fail in the same two ways: they are either a link tree with a background image, or a design project nobody can navigate. The middle is not complicated.

• A catalog page that lists everything you have released, each entry linking to every platform it is on. Not just the newest single.
• A way to actually listen, on the site, without an account. If someone arrives curious, do not send them away to satisfy that curiosity.
• An email capture that says what people are signing up for, and then only sends that.
• A press section — bio in three lengths, high-resolution photos, logo, contact address.
• A contact route that a human reads.

What does not belong: an intro animation, a mailing-list popup that fires before anyone has heard a note, and social feeds embedded in a way that makes the page slower than the platform they came from.

The maintenance question

The real reason artist sites die is not cost, it is friction. If publishing a release means editing HTML in three places, you will stop doing it by the third release. That is why this site keeps its whole catalog in a single data file: a new track is a few lines of JSON, and every page that mentions releases updates at once.

Whatever you build on, apply the same test before you commit: how many steps between “the song is out” and “the site says so”? If the answer is more than two, the site will be out of date within a year, and an out-of-date site is worse than no site at all.

The honest summary

A website will not break an artist. Nobody has ever gone from unknown to booked because their domain resolved. But every artist who lasts eventually needs a place that is theirs — for the press request, the sync opportunity, the label that wants to hear more than three singles, the fan who wants to go deeper than the algorithm allows.

Building that place is cheap and boring and takes a weekend. Not having it is only a problem on the day it matters, which is exactly the day you cannot fix it.

If you want to see the version of this argument taken to its conclusion, the player on this site is the whole catalog streaming without an account — the part most artist sites still send you elsewhere for.

The truth is duller and more useful. Nothing is decided in two days. But the first 48 hours do produce the signal that everything downstream is built on, and most of that signal comes from a small number of things you can actually influence.

What the platforms are watching

Recommendation systems differ in the details, but the questions they ask are broadly the same, and they are questions about behaviour, not volume:

• Did people finish it? Completion rate is the closest thing to a universal signal. A skip at twenty seconds says more than a thousand plays.
• Did they save it? A save is a deliberate act with a cost — it says “I want this again”, which is exactly what a recommender wants to predict.
• Did they come back? Repeat listens across days beat a single burst on day one.
• Did they add it to their own playlists? The single strongest endorsement an ordinary listener can give.

Notice what is missing: raw play count, follower count, and how many people you sent the link to. Those are inputs at best, and easy to fake, which is exactly why they carry so little weight.

What is worth doing

Given the above, the sensible release-day plan is short:

• Tell the people who already care, first and directly. Email list, close friends, the group chat. They are the ones who will finish and save it, which is the signal that matters.
• Give one clear instruction. “Save it” beats “check out my new song” — it is specific, it takes a second, and it is the exact behaviour the system reads.
• Make the link work everywhere. One smart link that opens in the listener's own platform, not a Spotify URL sent to somebody on Apple Music.
• Post more than once, differently. One announcement reaches a fraction of your audience. Three posts in different formats over two days is not spam, it is arithmetic.
• Pitch before, not after. Editorial pitching happens in the weeks before release. On the day itself, that window is closed.

What is not worth doing

Refreshing the stats page hourly. Buying anything that promises plays — the behavioural signal from purchased streams is uniformly terrible, and the systems are extremely good at spotting it. Emailing a hundred playlist curators you have never spoken to. Comparing your day-one numbers to an artist whose day one came after four years of touring.

Day one tells you who was already listening. Day thirty tells you whether the song is any good.

The thing nobody says out loud

The single biggest factor in how a release performs is the release itself. No promotion strategy rescues a track people do not finish, and a track people play twice needs remarkably little help. That is not a comforting conclusion when you have just spent a month on a rollout plan, but it is the one that survives contact with the data.

Which points at the actual strategy: release more often. Every release teaches you something, refreshes your presence, and gives the next one a slightly warmer start. Artists who put out six things a year learn six times faster than artists who put out one and promote it for a year.

After the 48 hours

The window that actually matters is the following month. Add the track to your own playlists. Send it to people individually as it comes up naturally. Write about it — a post explaining what a song is gives people a second reason to press play, and gives search engines something to index besides a title.

That last part is not incidental. Every release on this site gets written about in the journal, and the writing outlives the release-week noise by a considerable margin.

The player at sauvachi.com/listen is the other half. The full catalog, playable in the browser you are already in, with no account, no install prompt you have to accept, and no algorithm rearranging the running order.

What it does

• Plays the whole catalog — singles, collaborations and remixes, in the order they were meant to be heard.
• Keeps your place. Close the tab mid-track and it picks up where you left off, queue intact.
• Works offline. Save a release once and it stays on your device — trains, planes, and the parts of the city where signal goes to die.
• Installs like an app. Add it to your home screen and it opens full-screen, no browser bars, with playback controls on your lock screen.
• Remembers nothing about you. Saved tracks, volume, repeat and shuffle all live in your own browser storage. None of it is sent anywhere.

Putting it on your home screen

On iPhone: open /listen in Safari, tap the share button, then Add to Home Screen. On Android: open it in Chrome, tap the menu, then Install app. On desktop, most browsers show an install icon at the right-hand end of the address bar.

Either way you end up with an icon that launches straight into the catalog. It is the same page, minus everything around it — and because it is a web app rather than a store download, it updates itself the moment something new goes up. There is nothing to reinstall and no version to fall behind on.

Why build it at all

Streaming platforms are the distribution, and that is not changing — new releases go everywhere on day one, and always will. Most people will hear this music on Spotify or Apple Music, and that is genuinely fine. But the place with the artist's name on the door should be able to do more than list links to somewhere else. If you came here to listen, you should be able to listen here.

There is a practical side too. A link page tells you nothing about what a body of work sounds like as a body of work. Fifteen singles released over two years are a sequence, not a shelf, and the only place that sequence can be presented deliberately is a site nobody else owns.

The catalog belongs on the site with the artist's name on it. Everything else is a distribution channel.

Where to start

If you have never heard any of this before, the two most recent singles are the fastest way in — “Again” and “That Part”, both from July, and both linked below. From there the playlists are sequenced runs through the catalog rather than one long alphabetical dump, which is what the release list on its own would give you.

What is next

On the list: proper credits on every track, lyrics where they add something, and downloads for the people who still keep a library. Longer term, the goal is that a new release lands here at the same moment it lands everywhere else, with the writing that goes with it.

If something is broken or missing, the fastest way to reach me is Instagram. And if you would rather hear about new music by email than by algorithm, the subscribe button at the top of this page is the whole mailing list — announcements only, nothing else.

Where it came from

It started as a loop I could not leave alone — four bars that sounded finished the first night and wrong every night after. That is usually the sign to move on. This time the loop kept pulling things towards it instead: a drum pattern that sits further back than I normally place one, a vocal take recorded quietly because it was late, and a low end I rebuilt three times before it stopped fighting the sample.

The title came before the lyric did. Once a track is called “Again” it starts telling you what it is about, and the writing gets a lot faster. Repetition as a comfort and repetition as a trap are the same behaviour described by two different moods, and the song does not take a side.

What it sounds like

Slower than the last two singles and considerably darker. The palette is small on purpose: one sampled figure, one bass, drums, and a vocal that stays close to the microphone throughout. Nothing arrives to rescue the arrangement halfway through — if anything, things leave.

If you have heard “That Part”, this is the other end of the same month. That one is quick and blunt; this one takes its time and repeats itself until the repetition becomes the point.

How it was built

• The first take of the hook is the one on the record. Every cleaner version afterwards sounded like someone describing the feeling rather than having it.
• There is no second verse in the traditional sense. The section that would have been one is the same words, arranged differently — which is the entire point of the song.
• The drums sit late. Quantising them to the grid made the track sound urgent, and urgency is the wrong emotion for this one.
• It is deliberately short. It ends while you still want the next pass, because the alternative is a track that outstays a two-line idea.

Listen

“Again” is on Spotify, Apple Music and Tidal, and it is in the player on this site — you can hear it at sauvachi.com/listen without leaving the page or creating an account anywhere. The release card below has every link in one place.

If it lands with you

The useful things are the ordinary ones: save it, add it to a playlist that is not mine, send it to one person who would like it. That is genuinely how these travel — there is no marketing budget behind any of this, and a single share does more than it looks like it should.

More is finished than released at the moment, so the gaps should stay short. The journal is where each one gets written about properly, and the mailing list is the only place I say what is coming before it is out.

Two sittings

The first sitting produced the beat and the hook. The second one was three days later and consisted mostly of not adding anything. There is a version of this track with strings, a bridge and a second half that answers the first — none of it made the cut, because every addition made the song more polite, and the song is not supposed to be polite.

Working quickly is not a virtue on its own. What it does is stop you negotiating with an idea until it agrees with everything else you have ever made. Three days was long enough to check the thing was good and short enough that I could still remember why I liked it.

Most of the work on this one was deleting the work.

What to listen for

• The drop-out before the last chorus. Everything leaves except the vocal, for slightly longer than is comfortable. That gap is the reason the last section hits.
• The ad-libs are all first takes, recorded on the same pass as the lead. Comping them properly killed the momentum every time I tried.
• The hook lands early. No long intro to earn — the first thing you hear is the thing the song is about.
• No fade. It stops. Fades are for tracks that do not know how they end.

Where it sits

Next to “Again”, released six days later, this is the extrovert of the pair. Same month, same room, opposite instincts — one of them repeats an idea until it means something else, the other says it once and leaves. Played back to back in that order they make more sense than either does alone, which is roughly why both went out so close together.

Listen

Spotify, Apple Music and Tidal have it now, and it is streaming in the player on this site alongside the rest of the catalog — no account, no app to download. Every link is on the release card below.

There is more finished than released at the moment, so the gaps between these should stay short. The mailing list is the only place I say what is coming before it is out — the subscribe button at the top of the page is all there is to it.

How a remote collaboration actually goes

Nobody was in the same room at any point. The version that became the record is roughly the eleventh file in a shared folder, and the folder is a record of two people slowly converging: I sent a skeleton, psydfx sent it back with the harmonic content changed and the drums at half the tempo they started at, and the next four rounds were spent working out which of those two songs we were actually making.

Remote collaboration has a reputation for producing music that sounds assembled — two people politely leaving each other's parts alone. The fix, for us, was doing fewer rounds with bigger changes in each. Nobody nudging a hi-hat by three milliseconds and calling it a contribution; each pass had to justify itself by changing something that mattered.

What each of us brought

• The chord movement and the vocal are mine, more or less from the first sketch.
• The rhythm section and the sound design are psydfx's, entirely — including the texture running under the second half that I still could not tell you how to make.
• The arrangement is the thing we fought over, and the released order is psydfx's. It was the right call and it took me two weeks to admit it.

The bit that nearly did not survive

There is a moment about two-thirds through where everything drops to almost nothing before the last section. In my version it was four bars. In the version that shipped it is twice that, and it is the part of the track people mention first.

The lesson, appropriately enough, is that space is not empty — it is the thing that makes the loud part loud. I have a habit of filling gaps because a gap feels like an unfinished job. Watching someone else refuse to fill one, and being wrong about it in public for a fortnight, is a faster way to learn that than another year of doing it alone.

Working with someone whose instincts differ

The useful part of a collaboration is not the extra pair of hands. It is that every habit you cannot see in your own work is immediately visible to somebody else. Half the notes that came back on this track were about things I did not know I always do — start sections the same way, resolve tension too early, mix the vocal a little too far forward because I am used to hearing it that way.

None of that is a criticism of working alone, which is still how most of this catalog gets made. But a record that came out of an argument sounds different from one that came out of an agreement, and you can hear which this is.

Listen

“Lessons” is on Spotify, Apple Music and Tidal, and in the player here. All the links are on the card below. Go and hear what psydfx does on their own too — this record makes more sense once you have.

The idea

The concept was in the title before either verse existed: one track, two halves, and a beat that does not stay the same underneath them. NST takes the first half over the version we started with. I take the second half over what that beat turns into — same key, same tempo, almost nothing else in common.

Songs with two names on them usually sound like two guests visiting somebody's track: verse, verse, everyone goes home. This one is built so neither half works without the other. Play the second half on its own and it sounds like an answer to a question nobody asked, which is exactly the test we wanted it to fail.

The turn

Everything hinges on one bar in the middle. We tried it as a hard cut, as a filtered sweep, and as a drum fill; the version on the record does the least of the three, and it is the only one that did not feel like a trick. If you are listening on something with real low end, that bar is where the track changes its mind.

Getting there took longer than writing either verse. A transition is the one part of a song where the machinery is visible — do too much and the listener hears the edit instead of the moment. Doing almost nothing turned out to be the loudest option available.

Two verses, one beat, and one bar in the middle where it stops being the same song.

The video

Shot in a day and cut to the same structure as the track: the first half is static and close, the second half does not sit still for more than a couple of seconds. There is no story and nobody is acting. Cheap, fast, and it looks like the song sounds — which is the only brief that ever matters.

It is worth watching once with the visual and once without. The cut sells the turn so hard that the audio version can feel understated by comparison, and the audio version is the one that is actually correct.

Listen and watch

“THE SPLIT” is on Spotify, Apple Music and YouTube, and it is in the player on this site. Everything is linked on the release card below — the video is behind the YouTube button.

If you want the rest of the catalog in one place, the releases page has all of it, and the journal is where each new one gets written about on the day it lands.

This is the order I work in. Adjust it to taste, but do not reorder the first three.

1. Agree the splits before anything else

If more than one person touched the song, decide in writing who owns what before it goes anywhere near a distributor. Not after. Splits agreed while everyone is happy take five minutes; splits agreed after a song does well take lawyers.

Two separate things are being divided, and people confuse them constantly: the composition (the writing — melody, lyrics, chords) and the recording (the master itself). A producer might own half the master and a quarter of the composition. Write both down, get everyone to confirm by email at minimum, and keep it somewhere you will find it in three years.

2. Finalise the master

• Get a dedicated master, whether from an engineer or a service. Do not upload a loud mix and hope.
• Keep the highest-quality file you have — deliver WAV, 24-bit, at the session's sample rate. Distributors transcode; you should never be the one who compressed it first.
• Ask for an instrumental and a clean version at the same time. You will want both later, and going back to an engineer six months on is its own small ordeal.

3. Get the metadata right

This is the step everyone rushes and everyone regrets. Metadata is how your song is identified everywhere, forever, and some of it cannot be edited after release.

• Artist name spelled identically to your existing releases — capitalisation included. A different spelling creates a different artist profile.
• Features credited in the artist field, not buried in the title, if you want them to appear on the other artist's profile.
• Writer and producer credits filled in properly. This is what feeds the credits listing listeners see, and it costs nothing but attention.
• ISRC — usually issued by your distributor. One per recording, permanently. A remix or a re-record is a new recording and needs a new one.
• Explicit flag set honestly. Getting this wrong can pull a release from editorial consideration.

4. Artwork that survives being small

Requirements are consistent across the major platforms: square, 3000×3000 pixels, RGB, no URLs, no social handles, no “out now” text, nothing you do not have rights to. Beyond compliance, the only test that matters is what it looks like at 60 pixels in a queue — which is where nearly everyone will see it.

5. Pick a distributor and a date

Distributors differ mainly on cost model (per-release, annual, or commission) and on how quickly they pay out. For a first release the differences are smaller than the internet suggests. What matters more:

• Deliver at least three to four weeks ahead. Editorial pitching windows close well before release day, and delivering late forfeits them entirely.
• Friday release unless you have a specific reason not to — the industry's chart weeks are built around it.
• Keep your rights. Read what you are signing. Distribution should be a service you pay for, not an assignment of your masters.

6. Register everything

Distribution puts the song on platforms. It does not collect your publishing. Separately:

• Register the composition with your performing rights organisation.
• Make sure a publishing administrator is collecting mechanicals, unless you enjoy leaving money uncollected.
• Claim your artist profiles — the artist-facing dashboards on the major platforms — so you can update photos, bios and pitch tracks yourself.

7. The week before

• Check the release appears correctly on every store as pre-save links go live.
• Set up one smart link that routes each listener to their own platform.
• Prepare the assets you will actually post: a short clip, a still, the artwork, and the sentence you will say about it.
• Write the email to your list. Send it on the day, not three days later.

8. After release

Update your own site the same day — catalog entry, links, and the writing that goes with it. Add the track to your own playlists. Then start the next one: the most reliable promotion for a release is another release.

Two related pieces on this site go deeper on parts of the above: what actually moves in the first 48 hours, and why one smart link beats five platform links.

The title was a placeholder

Most of my files are named something useless — idea 4, new one, bounce final FINAL. This one got called “money gene” in the first session as a joke about a bassline that sounded expensive, and it stuck long enough to become true.

By the time there were lyrics, the title had already done half the writing. Songs are surprisingly obedient to their names. Call something “money gene” and it starts asking whether the appetite is inherited or learned, which is not a question I set out to write about and is now unmistakably what the song is about.

What it sounds like

Bigger and more open than most of the catalog. Where “CARBON” and the tracks around it are close and compressed, this one has air in it — the drums sit in a room rather than against your ear, and the low end is allowed to move instead of being pinned in place.

It is also the most straightforwardly enjoyable thing I have released. That is not an apology. There is a version of this that hides the hook under three layers of texture to prove something; that version was worse in every measurable way and I deleted it.

Notes from the session

• The bassline was written first, before drums and before any melodic content. Almost nothing else I have made starts there.
• The vocal was recorded twice: once properly, once quickly at the end of a session because the first one was too careful. The quick one is on the record.
• There is a full bar of near-silence before the last section. Everyone who heard the demo asked whether it was a glitch. It stayed.
• No key change, no dramatic outro. It stops where it stops.

Why the wide release

Every platform, deliberately. There is a temptation to focus effort on the one service where the numbers look best and let the rest happen automatically — but the people who use Boomplay or Amazon Music are not a rounding error, they are just somewhere else. Being missing from where someone already listens is the easiest audience loss there is to avoid, and it costs nothing to fix.

Listen

Every link is on the release card below, and the track is in the player on this site alongside the rest of the catalog — no account, no app.

If you want the mechanics behind releasing on that many services at once, the independent release checklist covers it end to end.

Roughly half the people who see it are not Spotify users. They tap, land on a page asking them to sign up for a service they do not want, and leave. The song was fine. The link was the problem.

What a smart link is

A smart link — you will also see them called pre-save links, landing pages, or by the brand names of the services that provide them — is one URL that shows the listener a page of every platform the release is on, and remembers which one they picked so it goes straight there next time.

That is the whole idea. One link in your bio, one link in the post, one link in the email, and every listener ends up in the app they already have.

What they actually solve

• Platform mismatch. The core problem, above.
• Link rot in bios. One URL you update in place, instead of editing five profiles every release.
• Pre-saves. Before release day, the same page can collect saves that fire automatically when the track goes live — which puts the song into people's libraries at the exact moment early behaviour is being measured.
• Measurement. You can see how many people clicked and which platforms they chose. Modest data, but more than a raw platform link gives you.

What they cost you

An extra tap. That is real, and it is why the advice is not “always use one”. If you are posting inside an environment where you know the platform — a Spotify-native context, or a message to one person you know uses Apple Music — link directly. Every unnecessary intermediate page loses some people.

There is also a dependency: your links now live on somebody else's service. If it disappears, every post you ever made points at nothing. Which is an argument for pointing at your own domain where you can, and letting your site handle the routing — the same reason the site itself matters.

Mistakes that make them worse than useless

• Listing dead platforms. A page with nine buttons where four go to stores nobody uses is a worse experience than three good ones.
• Leaving the default artwork and title. If the page does not immediately look like the song, people assume they mis-tapped.
• Gating the music behind an email form. Asking for an address before anyone has heard a note converts almost nobody and annoys everyone.
• Making a new link for every post. One canonical link per release. Fragmenting your own click data helps nobody.
• Forgetting it after release week. The link lives in bios and old posts for years. Check occasionally that it still resolves.

What we do here

Every entry in the catalog on this site carries both: the individual platform links and a single aggregate link, so the release page can serve someone who knows exactly what they want and someone who just wants to press play. The releases in the catalog each have their own page doing precisely that.

And for the case a smart link cannot solve — the person who does not want to leave at all — there is the player. No platform, no account, no redirect. Sometimes the best link is the one that does not send anyone anywhere.

What a remix owes the original

There are two schools. One says a remix is a service: keep the song recognisable, change the clothes, deliver it to a room the original could not reach. The other says a remix is a reading — you are allowed to disagree with the original, as long as the disagreement is interesting.

This is the second kind. The original moves one way and this version argues with it, and the vocal — untouched, in the same key, at a tempo it was not written for — has to carry the argument. That is the whole risk of the format. A vocal recorded to sit on top of one feeling has to sound intentional on top of a different one, and when it does not work, it does not slightly not work, it collapses.

What changed

• The tempo. Not dramatically, but enough that the vocal phrases land in different places against the beat.
• The harmony. The chords underneath imply something the original did not, which shifts what several lines seem to mean without a word being altered.
• The space. The original is dense. This is sparse for most of its length and dense exactly once.
• The ending. The original resolves. This one does not, on purpose.

Working with someone else's take

The strange thing about remixing a vocal you did not record is how much you learn about the singer. You hear every breath in isolation, every place they pushed and every place they held back, at a level of detail you never get from the finished record. By the end you know the performance better than you know most of your own.

It also enforces humility. You cannot ask the vocal to do anything different. Every problem has to be solved underneath it, which removes the easiest escape route in production — re-tracking until the arrangement is convenient — and leaves only the harder one, which is arranging properly.

A remix is the only format where the hardest constraint is the part you are not allowed to touch.

Listen

“To Kojayi (Remix)” is linked in full on the release card below, and it is in the player here. The other collaboration in this stretch of the catalog is “Lessons” with psydfx, which was built the opposite way round — two people changing everything, rather than one person changing everything but the voice.

Six things matter. In this order.

1. The room, not the microphone

The single biggest difference between a home vocal and a studio vocal is not the capsule, it is everything the sound hits after it leaves your mouth. A bedroom has parallel walls, a window, a wardrobe and a hard ceiling, and every one of them sends a slightly late copy of your voice back into the microphone.

You cannot fix that in the mix — reverb baked into a take is permanent. You can fix it with soft, thick things: a duvet on a stand behind the microphone, a wardrobe of clothes as the wall you face, a rug where the floor reflection lands. Cheap absorption in the right two places beats an expensive microphone in a bad room, every time.

2. Where you stand

Not in the middle of the room, and not facing a bare wall a metre away. Off-centre, facing into something soft, with distance behind you. Move two metres and re-record the same four bars — the difference is usually larger than any plugin you would reach for afterwards.

3. Distance and consistency

Close to the microphone gives you weight and intimacy; too close gives you plosives, proximity boom and a take that changes character every time you move your head. A hand's width is a reasonable default, with a pop filter to enforce it. The point is less the exact distance than holding it — most amateur takes are wildly inconsistent in level because the singer drifts, and no compressor fixes a performance recorded at four different distances.

4. Gain staging, unglamorously

Record with real headroom. Peaks around −12 to −6 dBFS are plenty; nothing is gained by pushing closer to zero, and a single clipped syllable in the best take of the night is an unrecoverable loss. Modern converters have more than enough range for you to be careful.

5. Headphone monitoring you can actually sing to

People sing better when they sound good to themselves. A little reverb in the monitoring path — not recorded, just in your ears — reliably produces more confident takes. Equally, if the beat is so loud you are shouting over it, you will get a strained performance and blame your voice. Turn the instrumental down.

6. Takes, not fixes

The most common home-recording mistake is treating editing as a substitute for performance. Tuning, timing correction and comping are all legitimate tools, but they are polish, not rescue. Four honest takes of a section beat forty repairs of one bad one, and they take less time.

Related: stop recording earlier than you think. Voices tire, and the take at hour three is almost never the one you keep. On “Again” the take on the record is the first one, recorded quietly because it was late — which is a pattern I have stopped fighting.

What actually needs buying

In order of return on money spent: absorption (blankets count), a stand and pop filter, a decent interface, then the microphone. That order is close to the reverse of how most people buy, which is why so many home setups have a costly microphone in a hopeless room.

None of this is about achieving a studio sound at home. It is about removing the specific artefacts that make a recording read as amateur — room reflections, inconsistent level, strained performance — so that what is left is just the song.

Writing forward from a standstill

There is a persistent idea that songs are reports — that you write what you are feeling as you feel it. Some are. But a lot of the most convincing music is aspirational rather than descriptive: you write the thing you are trying to talk yourself into.

This is that. Nothing in the weeks around it felt like momentum. The song is not lying about that, exactly; it is arguing with it. Which is why the production leans forward — a rhythm that pushes slightly ahead, a vocal that lands early rather than sitting back — while the words are considerably less certain than the arrangement is.

The build

• Everything enters early. There is no long introduction to earn: the first eight bars contain most of the elements you will hear.
• What changes over the song is density, not material. The same parts, more of them, arranged tighter.
• The vocal doubles only in the last section. Held back that long, doubling reads as a lift rather than a texture.
• No breakdown. It was written with one and it dragged.

Tempo is a mood, not a setting

The demo was slower by a noticeable margin, and every single thing wrong with it was fixed by speeding it up — which is not how it is supposed to work. Tempo gets treated as a technical parameter you settle early and leave alone, but it is closer to a key change in what it does to a song's meaning. The same words at 90 and at 100 are two different statements: one is resigned, the other is impatient.

The practical version of this: before you rewrite a section that is not working, try the whole track five BPM either side. It costs a minute and it solves a surprising proportion of arrangement problems that look like writing problems.

Where it sits in the catalog

Between the close, compressed tracks and the more open ones. If you want the version of this energy with the brakes on, “Autumn Leaves” is the closest neighbour; if you want it with the brakes removed entirely, that is “money gene”.

It is also the track I would point a new listener at first. Not because it is the best one, but because it is the most representative — everything I tend to do is in it, and none of it is hidden behind a production choice you have to acclimatise to.

Listen

All the links are on the release card below, and it is in the player on this site. The catalog page has everything else in one place.

Here is the structure underneath it. None of this is advice — it is the shape of the system, and you should check every detail against your own distributor and society statements, because the specifics vary by country, service and deal.

There is no per-stream rate

The most important thing to understand: the major services do not pay a fixed amount per play. They operate a revenue pool. Subscription and advertising income for a market is collected, a share goes to the service, and the rest is divided among rights holders according to their share of total streams in that market.

Which means your effective rate moves depending on things that have nothing to do with you: how many people streamed anything at all that month, what proportion of your listeners were on free tiers, which countries they were in, and what the exchange rate did. The commonly quoted figures — fractions of a cent per stream — are averages after the fact, not prices.

Two different royalties, always

Every stream generates money on two separate sides, and they travel by completely different routes:

• The recording (master) side. Paid to whoever owns the recording. If you are independent and distributing yourself, that is you, and it arrives via your distributor. This is the larger share.
• The publishing (composition) side. Paid to the writers and publishers of the underlying song. It arrives through your performing rights organisation and a publishing administrator — not through your distributor.

This is where most independent money goes uncollected. An artist who distributes but never registers the composition is leaving the entire publishing side sitting in a collection society's unmatched pile. It is not a large share of a small number, but it is your share of a small number, and it compounds across a catalog.

What actually moves the total

Given the pool model, three things matter far more than chasing a better rate:

• Catalog size. Income is roughly the sum of many small streams of income. Twenty releases earning modestly outperform one release earning well, and they keep earning while you sleep.
• Repeat listening. A listener who plays a track fifty times over a year is worth vastly more than fifty listeners who play it once, and is also far more likely to buy something.
• Everything that is not streaming. Merchandise, live, sync licensing, direct sales. For most independent artists these dwarf streaming income, and they are the reason streaming is best understood as distribution and discovery rather than as revenue.

Streaming is how people find the music. It is rarely how the music gets paid for.

The collection checklist

If you do nothing else, do these:

• Register every composition with your performing rights organisation, including the ones that seem too small to bother with.
• Get a publishing administrator collecting mechanicals internationally.
• Claim your artist profiles on each platform so the credits and metadata are right — mismatched metadata is a common cause of money failing to match to you.
• Keep your splits documented. Uncertain splits are the single most common reason payments stall.
• Read one statement properly, all the way through, once. It is tedious and it will teach you more than any article, this one included.

The honest framing

Streaming pays badly per play and will keep doing so, because the pool is finite and the number of streams keeps rising. Treating it as a wage is a route to permanent disappointment. Treating it as the world's cheapest distribution — a way for anyone, anywhere, to hear your catalog at no cost to you — is both accurate and considerably more useful.

The practical response is the same either way: release more, keep your rights, collect what you are owed, and own the places your audience can reach you directly. That last one is why a site and a mailing list still matter.

A song that refuses to build

Every instinct in modern production pushes towards escalation: add a layer, lift the chorus, double the drums, give the last section something the others did not have. Those instincts are correct most of the time, which is exactly why a song that declines to obey them is difficult to finish. You keep reaching for the lift, and every time you add it, the song gets less interesting.

This one ends roughly where it starts. There are four versions in the project folder with a proper climax in them and all four are worse. Eventually the discipline became the point: hold the level, let the writing carry the change, and trust the listener to stay without being shouted at.

What is actually in it

• Very little. A sparse rhythmic bed, one harmonic idea, and a vocal that stays in a narrow range for almost the entire song.
• What movement there is comes from removal — parts leave and do not return.
• The reverb is short and dark rather than long and pretty; long tails would have made it sound like a ballad, which it is not.
• The last line is unaccompanied. Everything else exists to make that possible.

Restraint is a skill you have to practise

Doing less is talked about as though it were the easy option. It is not — it is a technique, and it takes as much work as density does, just less visibly. Sparse arrangements expose everything: timing, tuning, the room, the decision to breathe where you breathed. There is nowhere for a weak part to hide, so every part has to be right.

Making this taught me more about mixing than anything loud ever has. When there are four elements, the relationship between them is the whole record.

It also changes what mistakes cost. In a dense arrangement, a slightly late note is absorbed by everything around it; here it is the only thing happening, so it reads as an error rather than as feel. Several takes were discarded for reasons that would have been completely inaudible on any other track in the catalog.

The title, and what it is not about

It is not a cover, and it is not really about autumn. The phrase arrived attached to the melody, the way titles usually do, and by the time I noticed it carried a hundred years of other people's associations it had already become the only possible name for it.

What the song is about is closer to the mechanics of that: how something ordinary and repeated — a season, a habit, a person leaving in the same way they left last time — stops being an event and becomes a schedule. Which is also why it declines to build. Schedules do not have climaxes.

Listen

The links are on the card below, and it is in the player. If you want to hear the opposite instinct from the same catalog, “money gene” is as far the other way as this goes.

Starting with nothing to say

The received wisdom is to come to a session with ideas — a concept, a hook, at least a direction. Good advice, generally. It is also how you end up with two people politely taking turns to execute two separate songs on the same beat.

We started from an instrumental and a phrase neither of us meant seriously. “Can't complain” is what you say when you do not want to answer the question — a deflection so ordinary that nobody hears it any more. Written down and repeated a few times, it stops being a shrug and starts sounding like something much less relaxed, which is the entire song.

How two voices divide a track

Most features are structural: your verse, my verse, everyone home by the third chorus. This is closer to a conversation. The sections trade faster than that, and there are places where the two vocals overlap deliberately rather than tidily handing off — which took a long time to mix, because two people talking at once is either a texture or a mess depending on decisions of a few decibels.

• The lead alternates, but the ad-libs are always the other person. It keeps both voices present even when only one is singing.
• The hook is sung together, in unison rather than harmony — harmony made it sound like a duet, and it is not one.
• Neither of us gets the last word. It ends mid-exchange.

What I took from it

That the useful part of a collaboration is being asked to justify things. Working alone, most decisions are made and forgotten in the same minute. With someone else in the room, every one of them briefly has to be defensible — and about a third of mine turned out not to be.

The other thing, less comfortable: two writers on one instrumental will always converge on the more obvious of the two available songs unless somebody actively resists it. Agreement feels like progress and often is not. The sections here that I like most are the ones where we did not agree quickly.

Splits, and having the boring conversation early

We agreed the writing and master splits before the second session, in a message thread, in plain language. It took about ten minutes and it is the reason this release exists rather than sitting unreleased in a folder — which is where a genuinely large number of good collaborations end up.

Nobody wants to have that conversation while the song is still fun. Have it anyway: the alternative is having it later, when there is money or credit attached and the friendship is doing the negotiating. There is more on how splits actually work in the release checklist.

The other collaborations in the catalog work differently again: “Lessons” with psydfx was two people rebuilding each other's work remotely, and “THE SPLIT” with NST is deliberately two halves that refuse to merge.

Listen

Every link is on the release card below, and the track is in the player here.

Here is what is actually happening in that window.

Week three: delivery

You upload the master, the artwork and the metadata to a distributor. What they do next is convert your submission into the format the stores accept — a delivery package containing the audio, the artwork, and a large amount of structured data describing who made what.

Two things go wrong here more than anything else. The first is metadata: a name spelled differently from your previous releases creates a second artist profile, and untangling that afterwards is a support ticket, not an edit. The second is artwork rejection — text that reads as promotional, a logo you do not have rights to, or a resolution below requirement.

Week three, immediately after: the pitching window

This is the part with a hard deadline. Editorial playlist consideration on the major platforms happens through the artist-facing dashboards before a track is live, and once the release date passes, the opportunity is simply gone. Not reduced — gone.

Which is why the standard advice is to deliver three to four weeks ahead. It is not the distributor that needs the time. It is you, so that the release exists in the system early enough to be pitched, pre-saved and checked.

Weeks two and one: store ingestion

Each store takes the package and processes it independently. They transcode your audio into their own formats, generate artwork at a dozen sizes, run automated checks, and — at some services — put the release through review.

Because this is per-store, releases can and do arrive in slightly different states in different places. One platform shows the feature in the title, another in the artist field. One has the artwork, another shows a placeholder for a day. One goes live at midnight local, another at midnight in a different timezone entirely. None of that is a bug on your end; it is a dozen pipelines with different rules running the same package.

The pre-release checks worth doing

• Look at every store page as soon as pre-save links appear. Wrong artwork or a mis-attributed feature is fixable before release and painful after.
• Play the preview clip. Some stores auto-select it; occasionally they pick something unrepresentative.
• Check your artist profile is the one the release attached to, not a new empty duplicate.
• Confirm the explicit flag matches reality in both directions.
• Test your smart link on a phone that is not yours.

Release day itself

Almost nothing happens mechanically. The stores flip the release visible on the date you specified, in each territory's local time, which is why your track can be live in one country hours before another. If something is broken at this point, the fix goes back through your distributor and takes as long as ingestion took the first time — usually a day or two, sometimes longer.

Why this matters more than it sounds

Every avoidable problem in this window costs the same thing: the release-week attention you only get once. A track that goes live with the wrong artwork on the biggest platform, on the one day people are actually looking, does not get a second launch.

The full sequence, including everything before delivery, is in the independent release checklist.

The folder

Everyone who makes music has one: dozens of projects that stopped somewhere between an idea and a song. Most producers treat it as a graveyard. It is closer to a warehouse — the reason those ideas stalled is usually not that they were bad, but that they arrived on a day when nothing around them was ready.

Going back through it is uncomfortable in a specific way. You meet a version of yourself who had a good idea and no ability to finish it, and the honest response is not embarrassment but salvage.

What survives the return trip

• Loops survive. Anything with a strong four-bar identity still works years later.
• Arrangements do not. Every structural decision in those old files was wrong or irrelevant, without exception.
• Sounds survive; processing chains do not. The source material was worth keeping, the twelve plugins on top of it were not.
• Vocals rarely survive. A performance is tied to what the person meant at the time, and it does not transplant cleanly.

So the method became: keep the four bars, throw away everything that ever happened after them, and write forward from there as if it were new. Which it effectively was.

Why it is worth doing at all

Because the alternative — always starting from nothing — quietly biases what you make towards whatever you are capable of in a single sitting. An old idea has no obligation to sound like the person you are this month. Half the appeal of this track is that it does not sound like the tracks around it, and the reason for that is simply that its raw material predates them.

Unfinished ideas are not failures. They are material you have already paid for.

How to keep a folder worth returning to

The salvage only works if the archive is navigable, and most are not. Three habits make the difference, none of them interesting:

• Bounce a rough audio file of every session you abandon. In two years the project may not open — plugins expire, software changes, licences lapse — but a WAV always will.
• Name it after what it sounds like, not the date. “slow chords wrong key” is findable; “Untitled 47” is not.
• Write one line about why you stopped. “Bored of the drums” and “could not sing it” send you back with completely different intentions.

That is roughly ninety seconds of work per abandoned idea, and it is the difference between an archive and a landfill.

Listen

Links on the release card below, and it is in the player here with everything else.

Rebuilding is usually a symptom

There is a specific trap in home production: when a track feels not-quite-right, the available action is always more work, so that is what you do. Change the drums. Re-record the vocal. Try it eight BPM faster. Each pass feels productive and the song gets steadily further from whatever made you start it.

What is usually wrong at that point is not the production. It is that a section is a bar too long, or the vocal is sitting in the wrong place emotionally, or you are bored of a song you have heard four hundred times and mistaking that for a fault in the song.

How you tell the difference

• Play it to someone who has never heard it. You will hear it through their ears for two minutes, which is worth a week of solo listening.
• Leave it for two weeks, untouched. Problems that are real are still obvious on return; problems that were fatigue evaporate.
• Check whether you can name the fault. “The second verse drags” is actionable. “It just needs something” means stop.
• Keep every version. The only reason this record exists is that the original demo was still in the folder after two rebuilds had wandered off.

What the two of us actually changed

In the end: the arrangement of the last third, the drum sound, and roughly four lines. That is a normal amount of work for a song. It took three months because we spent two of them doing considerably more than that and then undoing it.

Sadra was also the person who said, halfway through the second rebuild, that we had already made the record and were now making a different, worse one. That is the sort of thing that is very hard to say to yourself and very easy to hear from someone else — the same argument for collaboration as in “Can't Complain”.

The version-keeping habit

Save a dated copy at every point the track is in a state you would be willing to release. Not at every save — at every coherent state. It costs disk space, which is free, and it buys you the ability to be wrong for a month without permanent consequences.

Almost nobody does this, and the reason is psychological rather than technical: keeping an old version feels like an admission that the work you are about to do might make things worse. It might. That is exactly why you keep it. Three of the tracks in this catalog exist because a superseded bounce was still sitting in the folder when we came to our senses.

Listen

All links on the release card below, plus the player on this site.

Assume ten seconds of patience. Design accordingly.

The six things everyone needs

• A one-line description. Who you are and what you sound like, in a sentence someone could paste into a listing without editing.
• A bio in three lengths — roughly 50, 150 and 400 words. Different outlets need different sizes and none of them want to cut yours down.
• Photos that can actually be printed. High resolution, at least one horizontal and one vertical, with the photographer's credit attached.
• Music, playable immediately. A link that plays without an account. If a curator has to sign up for something to hear you, you have lost.
• Contact. A working email address, on the page. Not a form, not a DM request.
• The factual block — release dates, label or self-released, territory, and the exact spelling of every name involved.

What quietly ruins it

• A download that is 400 MB of raw photos. Nobody is downloading that on a deadline.
• A bio written in the third person about your “journey”. Describe the music. Three specifics beat a paragraph of adjectives.
• Comparisons to enormous artists. They read as a claim rather than a description, and everyone discounts them automatically.
• Unlabelled photos. If a publication cannot tell who took the picture, they cannot legally run it.
• Numbers with no context. A stream count is only persuasive if it is genuinely persuasive.
• A press kit that is out of date. A kit whose newest release is two years old says more than the kit intends to.

The format question

A PDF is still the most portable option — it opens everywhere, prints, and can be attached to an email. A web page is easier to keep current and easier to link. The pragmatic answer is both: a page that is always right, and a PDF generated from it for people who want a file.

Whichever you pick, the important part is that it is one link you can send without preamble, and that it lives on a domain you control. This site keeps its press kit reachable from every page's header for exactly that reason, alongside the argument for owning the address in the first place.

The test

Send it to a friend who does not work in music and ask them to find your email address, your newest release, and a printable photo. Time them. If it takes longer than a minute, the problem is not that they were slow.

Claustrophobia as a production choice

Most mixing advice is about creating space: carve room for each element, keep the low end tidy, let the vocal breathe. Good advice. This track ignores nearly all of it on purpose.

Everything sits in the same narrow band. The vocal is pushed hard against the instrumental rather than floated above it. There is almost no stereo width, so the whole thing arrives from one place, close in. It is a deliberately uncomfortable listen for two and a half minutes, and it would be unbearable at four.

Loud is not big

The counter-intuitive part, which took me years to internalise: heavily compressed music does not sound powerful, it sounds tense. Power comes from contrast — the loud part is loud because the thing before it was not. Remove the dynamic range and you remove the mechanism that makes impact possible.

So the trick with a track like this is to spend that tension rather than just maintaining it. There is exactly one moment where everything opens, and the entire arrangement exists to make that moment land. Play it after something spacious like “Autumn Leaves” and the effect roughly doubles.

Notes

• Mixed almost entirely at low volume. Anything that survives being quiet will survive being loud; the reverse is not true.
• No reverb on the lead vocal at all — only a very short room, which reads as tightness rather than as an effect.
• The master is deliberately not as loud as it could be. Pushed further, the one open moment stopped opening.

Short on purpose

Two and a half minutes is not a concession to attention spans, whatever the current argument about song lengths says. It is a structural requirement. A track built on sustained tension has a natural expiry: past a certain point the listener's ear adapts, the discomfort stops registering, and what was claustrophobic becomes merely loud.

You can hear where that line is if you loop the thing for ten minutes. Somewhere around the third pass it stops working entirely. Ending before that happens is not editing for the sake of brevity — it is the only way the idea survives being played twice.

What I would do differently

Very little, which is unusual. The one thing: the vocal could sit a decibel further back in the second half. It is mixed at the level that felt right while I was making it, and making it involved hearing the words several hundred more times than any listener will. That is a bias worth naming, because it recurs — the person who wrote the lyric always mixes it slightly too loud.

Listen

Links on the release card below, and it is in the player here. Headphones, ideally, and once through without doing anything else.

Failure mode one: the low end

Headphones deliver bass directly to your ears with none of the cancellation a room introduces, which usually means you hear more low end than a listener will on speakers — so you under-provide it — while simultaneously being unable to hear whether a sub note is actually in tune or in time.

The checks:

• Reference against two or three commercial tracks you know intimately, matched for loudness. Trust the comparison, not your ears in isolation.
• Check on something small — a laptop, a phone speaker. If the track collapses into nothing, your bass is carrying too much of the arrangement's weight.
• Look at the low end, do not only listen: a spectrum analyser will show you a build-up your headphones flatter.

Failure mode two: width and reverb

Headphones exaggerate stereo dramatically, because each ear hears only its own channel. Wide sounds spectacular on headphones and can nearly vanish on a phone. Reverb behaves similarly: the tail you carefully placed can be inaudible in a car, or turn to soup in a room.

• Check the whole mix in mono, regularly. Anything that disappears was phase, not width.
• Keep the important elements — lead vocal, kick, bass, snare — centred and dry enough to survive mono.
• Judge reverb amounts at low volume. Quiet listening is where too much reverb is most obvious.

Failure mode three: your own fatigue

Headphones are exhausting in a way speakers are not, and tired ears reliably reach for more brightness and more level. Two hours in, you are no longer mixing, you are gradually turning everything up.

• Mix quietly. Genuinely quietly — quieter than feels satisfying.
• Take breaks on a fixed schedule rather than when you notice fatigue, because by then it has been affecting decisions for a while.
• Make the final call the next morning. Every mix sounds different at 9am, and 9am is right.

What actually helps

In order: knowing your headphones well (one familiar pair beats three unfamiliar ones), referencing against known records, checking in mono and on a phone, and mixing at low volume. Correction software that flattens your specific model's response is genuinely useful and cheap; a second pair of expensive headphones is much less so.

None of this makes headphones equivalent to a treated room. It makes their specific distortions predictable, and a predictable distortion is one you can mix through. The rest is the same discipline as recording in an untreated room: fix what is cheap to fix, learn what the rest is doing to you, and stop pretending the problem is gear.