Building the Session

Short, dry, slightly lo-fi samples with no room on them

Load the samples dry, because the room comes from the returns. The track sends to the plate at -6 dB, and the Drums bus it feeds sends to the hall at -18 dB. That's 6 dB less than the guitars' bus sends, so the drums share the guitars' room and sit further back in it. A short, cheap sample with nothing on it is still what this genre's kit sounds like.

Don’t add swing or timing humanisation. The part sits dead on the grid on purpose, while the electric bass, the guitars, the keys, the pad, the lead and the acoustic kit drift around it the way hands do. That contrast is the sound. If the kit feels stiff, reach for tape saturation or a little bit-reduction instead.

Velocity here is a programmed accent. The kick and the closed hat hit harder on the beat than off it. The snare and clap always play at their accented level, and the open hat, toms and crash each keep one fixed level. A fill varies its strokes and a snare roll climbs, and that's all that moves. A fuller section plays a busier groove, never a louder one.

A close-mic’d, dead kit with no room in the samples

Dry samples let the room come from your mix rather than from wherever the kit was recorded. Which kit you load and how the part is played are separate choices, so pick the kit by ear.

Re-read the note map before you swap kits. The part is written to a map close to General MIDI without being it. The rimshot sits on 40, and the open hat runs 12 to 17 as a continuum rather than one note, so a kit mapped straight to GM loses both. Nothing here writes a hand clap.

This track and DRM Machine play the same groove. Mute one. Keeping both gives you a layered kit, which is a real sound and a different record. This kit drifts a few ticks around the machine kit, so with both up the kicks flam a little, and tightening this one to the grid runs them together. In any bar this kit plays, the machine leaves the crash to it. Where this kit sits out, the machine keeps its own. So if you mute this one and keep the machine, copy its crashes across or those bars go without a cymbal. The ending's last kick and crash go on whichever kits play the final bar.

A shaker, a tambourine or claps, one piece per song

Short hits on whichever piece the song picked. The shaker sits on 126, the tambourine on 54 and the claps on 39, so check your kit's map. A song where the drummer's on the ride never gets a tambourine.

It keeps the time where the kit drops out. A shaker or tambourine carries on through the bars the drums leave, so the pulse doesn't go with them.

Claps only play two and four. And only in the choruses and the outro.

Each section plays its own figure. It's the kind of figure this genre's percussion actually plays, and it gets busier as the section builds. The hits sit a few ticks off the grid, the way a hand plays them, and every take stays at one weight.

Electric or synth, and each song is written for one of them

An electric bass wants a sampled four-string under it. A synth bass wants something like a Minimoog-style mono or a Juno bass.

The two are written differently. An electric bass sits about four ticks off the grid, the way a session bassist plays. A synth bass arrives sequenced, on the grid at one level, so there's nothing for you to flatten.

Sampled basses disagree about octaves. These notes are written so C3 is 60 and a four-string's open low E is 40, and nothing goes below it. An instrument that calls that low E 28 plays the whole part an octave high, so transpose the track down one. Nothing's wrong with the MIDI. The two conventions just disagree.

The Live set adds mutes and ghosts. The MIDI file doesn’t. On about a quarter of sampled-bass songs a few notes sit far below the playable range, at D#-1 and E-1. Those are keyswitches, each one a sixteenth before the root on the next downbeat. Both restrike the note before it softer, one as a mute and one as a ghost. On any other bass, mute or delete them. They stay out of the plain MIDI on purpose, because anything that ignores keyswitches plays them as real low notes.

Clean single-coil, played with a plectrum

A Telecaster with both pickups on, or a Rickenbacker if you have one, which is what this part is reaching for. A 12-string works too. Play it with a plectrum, because jangle is a transient and fingers round off the thing that makes this part ring. A Strat is glassier than this wants.

Into a Fender Twin or a Roland JC-120, whose own chorus is the classic clean sound here. Chorus, then an analogue delay. The Jangle bus sends it to the plate, so switch the amp's reverb off.

Swap the chorus for amp tremolo and the same figure turns noir. Darker than sunlit. It suits a slow figure with space in it, where on a busy one the tremolo and the picking fight over the pulse. Only if the Rhodes isn't already tremoloing, because that's where the effect lives by default.

No overdrive. This genre’s guitar is undistorted and treble-forward. The one exception is the gain-led end of it, where fuzz is a texture wash rather than a driven tone.

A real second take, not a copy

The notes are identical to the Guitar track. The timing and velocity of every note are not, and that difference is the whole effect. Two identical takes sum back to the centre however you pan them, so a copy only makes the part louder. Each note here sits 10 to 25ms off its twin, which is the window where it widens. Tighter than about 10ms the two comb, and by 40ms you hear a slapback instead of one wide guitar.

Pan hard. Guitar 100% left, Guitar 2 100% right, equal level. Any closer and each speaker gets both takes, which combs.

Give the two sides different tone. Two live takes differ in pick position, string excitation and amp state. Two MIDI tracks through one instrument don't, so tone is the one thing the MIDI can't supply. Put a separate amp or cab on each side and change something real. A different cabinet, a moved mic, a few dB of EQ tilt in opposite directions.

Round-robins do a lot of the work. On a library that round-robins its samples, the velocities written into this track pull a different sample on each side, which is variation a real double gets for free. Leave them alone rather than flattening them before the instrument.

Detune one side very slightly if you want more. About 7 cents, in the instrument or with a pitch utility, where you can hear what you are doing. It is deliberately not in the MIDI, because pitch bend depends on each instrument's own bend range and 7 cents on one plugin is 40 on another. Keep it small. Heavy detuning on top of a time offset is what makes doubles sound flabby.

It plays in the choruses. Only the ones the Guitar plays in. Everywhere else the Guitar already plays alone, still hard left with its delay send up.

Rhythm only, by design. There’s no Lead double and there shouldn’t be. Doubling smears the one part that needs to stay singular and centred, so widen the lead line with a harmony or a delay instead.

A Rhodes, soft and bell-like rather than barked

A Suitcase or a Mark I. A Wurlitzer is the grittier alternative where a track wants more bite.

This is tremolo’s first home. The Rhodes' own amp tremolo, slow and shallow, is one of this genre's signature textures, and here it's the default rather than a turn toward the noir end. One instrument only.

The voicings leave the root to the bass. In a bar with no bass under it the comp puts the root back. So a bright, mid-forward patch beats a full-range piano sound.

It comps beside the guitar and the arp. Its figure is the kind this genre’s keyboardists play, on the same eighths or sixteenths as the kit. Every voicing puts its top note above the guitar’s highest, so both can move at once and the Rhodes still sits clear of the picking.

A Juno-style poly, low-passed and sitting under everything

The Juno-106’s built-in chorus is most of this genre’s pad sound on its own. A Prophet-style poly is the warmer alternative.

Slow attack, long release. Chords have to bleed into each other. The part strikes each chord and holds it until the next one, letting go just before that next attack, so a fast envelope leaves a gap where the bleed should be.

Then big reverb, and don't be shy with it. This is the track carrying the haze the whole style is named for.

A warm pad or string patch with a very slow attack

Heavy reverb, and a long tail straight across the section boundary, because the gesture is what glues a section change together.

The envelope is the gesture. What arrives is one chord held about three bars, a touch shorter at a slow tempo and out to four at a fast one. The swell is the patch's slow attack opening across the first bar and its long release falling away over the rest. On a patch that speaks instantly it plays as a flat block chord, so lengthen the attack and release rather than looking for missing notes. A matching curve is written to CC7 and CC11 for hosts that read them.

It’s meant to be sparse. One chord on each handover into a thicker section, and silence in between. If it feels absent, turn it up rather than asking for more of them.

Thin and bright rather than fat

A narrow pulse or square, short decay, minimal sustain, so each note is a blip for the delay to smear rather than a held tone. Cheap and slightly out of tune is a feature here, and it's why the part shimmers instead of sounding clinical.

The movement has to come from effects. The part sits dead on the grid, because a real arpeggiator is a sequencer, and only the first step of each bar is struck harder. A sparse section keeps it inside one octave and every other section runs the figure's own span, with a full chorus adding a second arpeggiator an octave up, a touch softer. What makes it move is a synced delay and a slowly swept filter.

When it sits on one note, that’s on purpose. Where the guitar is picking, the arp keeps its rhythm and stays on one note of the chord, because the guitar’s already breaking the chord up. Everywhere else it runs the full arpeggio.

Clean single-coil, pushed back in space

A Stratocaster on the neck pickup, played with fingers, into a Fender Twin or a Roland JC-120.

It is the opposite of the picked part on purpose. Neck against bridge, fingers against a plectrum, one voice against four. Those three differences carry the separation, which is why both parts can come out of the same library. A rounded attack also stays out of the singer's way, where a pick's transient argues with the consonants.

Chorus and flanger both belong to the amp. Take them from the amp suite's own pedals rather than putting them after it, so they sit inside the tone. Chorus goes on all three guitars. A slow flanger is the sound of a dream pop lead line, so put it on the Lead when a song wants one and leave it off everywhere else.

Amp tremolo is the other way to take it. It gives you the twanging, noir lead the desert end of this sound is built on. One tremolo per song though, and the Rhodes has first claim.

On a long sustained contour, an EBow or a volume pedal gets you a bowed, attackless note rather than a picked one. If it needs more body, layer a second quiet take instead of reaching for gain.

On a sampled guitar, use melody mode with legato enabled. Every lead note sits on a string and a fret, under a hand that stays in one position where it can. Two notes in a row on one string inside that position are a hammer-on or a pull-off of up to four frets, and a note on the next string over is picked. So the first overlaps the second by a few dozen ticks, and the second's written softer because nothing picks it, which on some libraries is what calls up the hammer-on sample. Set to re-pick every note, a library plays the pair as two picks.

Set the bend range to 2. The lead writes its pitch wheel for two semitones either way, so any other setting makes its slides miss their notes. A slide comes up from a fret or two below, either a shift along the string inside a run or the way into a long note after a rest. It never bends a string. In this genre the pitch moves on the tremolo arm and through the pedals, and that's the mix's job rather than the MIDI's.

From the second chorus it plays a counter-melody. Until then it loops a short figure through every bar the singer sings, a little hook with a rhythm of its own, and plays a phrase wherever the singer stops for two whole bars. After that it's a second tune underneath, between the bass and the vocal line. It moves where the vocal holds or rests, holds through the busy passages, and never lands on the singer's note. A low voice often leaves no room underneath, so it keeps looping. In the last chorus, where the singer's hook fits its range, it plays that hook an octave up. With the Premium plan's Lead under the singer switch off on the write page, it keeps looping right through and still plays that last hook.

It never takes a solo. In an instrumental after the second chorus it keeps its own figure going right through, instead of climbing to a high point. Dream pop guitarists build parts out of chords and arpeggios, not solos.

The tune, played rather than sung, so keep it deliberately plain

A blown instrument is the closest stand-in. It covers a voice's range, plays one note at a time, and every note starts with breath and can swell or fall away. Keep the tone dark enough to sit where a voice sits rather than over the top of it.

A bowed cello is the other good answer. Bow pressure moves across a line the way breath does. The trade is a vaguer attack, and a clear attack is what you want while you're checking a tune against a set of words.

A plain oscillator is the stricter version. One sine or triangle, around 20ms attack and 200ms release, a little vibrato, portamento off. The pitch is unambiguous and it won't flatter a line that doesn't work.

Resist choir and vox patches. They smear pitch and blend into the Pad, which hides the exact problem you're listening for.

Learn the verse and the chorus separately. The chorus is written on its own chords rather than lifted off the verse, and it repeats harder, which is most of what makes it a chorus.

The line is written against the chords, so it transposes with the song. Where the range doesn't suit your singer, move the whole track rather than editing notes.

The stacked chorus, and only some songs have one

It loads whatever the vocal above it loads. The pan and the level are what make it read as a second voice, not a different sound. It gets its own track so you can balance the two.

It only sings where the singer states the hook in a chorus, a third or a sixth off the tune. Most songs put it above and the rest below, and it moves with the vocal step for step through the scale, keeping one interval for a whole phrase. It leaves that interval only where the finished line would clash with the band. Then a note on one or three, or a held note, takes a chord tone that's still a third or sixth off the tune, and the phrase's last note ends on the chord.

Pan it a little off the lead vocal and sit it under rather than beside, which is why it arrives a couple of decibels quieter.

The vocal line doubled in the last chorus

It loads whatever the two vocal tracks above it load. Sung by a person, a second take never quite matches the first, and that's the sound. A sampled voice or a vocal synth plays the same notes the same way and mostly just comes out louder, so detune this track a few cents or slide it a few milliseconds late.

It's the singer's line sung a second time, the same notes on the same timing, the way Cocteau Twins double their lead. It sits a little off the lead on the side away from the harmony, and a touch softer than the original.

It’s the last layer on the singer. The harmony comes in first and the double last, so it sings the last chorus and nothing else. A song with one chorus has no double, and neither does one whose harmony waits for its last chorus. The rest all get one.