Chapter 9 Key Takeaways
What you should walk away with
- MIDI is a script, not a recording. A MIDI clip contains actions — this pitch, struck this hard, now — and no sound at all. The sound happens when a virtual instrument performs the script, which is why you can change the notes, the tempo, or the entire instrument tomorrow without losing a thing. The bill for that freedom: nothing is ever committed, and producers have died of fiddling. Decide when a part is done, and mean it.
- The piano roll is for reading, not just entering. Pitch runs vertical, time runs horizontal, and note length is a musical decision, not an accident of where the rectangle got drawn. Once you're fluent, you can see a muddy low-voiced chord, a flatlined melody, or an overcrowded arrangement before you press play. Beginners obsess over note starts; players-on-purpose shape note ends, because that's where phrases breathe.
- Velocity is THE expression channel. It isn't a volume knob per note — on any decent instrument it selects timbre, because real instruments get brighter when struck harder and your ear has known that correlation its whole life. Flat velocity is the loudest amateur tell in all of programming: one timbre, rubber-stamped. Arcs toward phrase peaks, top-note hierarchy in chords, repeats that never match — that pass removes the stamp in about thirty seconds per part.
- Quantize has a strength slider, and the slider is the craft. 100% teleports a take to dead; 50–80% tightens a performance while keeping its push and pull. Quantize note starts, almost never note ends. Swing gets a taste here and a chapter of its own later (Chapter 13) — apply it by ear, in small amounts. The grid is a reference, not a residence.
- Humanization means intention, not randomness. The randomize button adds jitter; real players add pattern — chords rolled the same direction every time, a bass that leans late all song, phrases that shorten into their exits, repetition that never repeats exactly. Your brain forecasts the beat and reads structured deviation as a person and random deviation as a malfunction. Five to twenty milliseconds, with a reason every time.
- Samplers are recorded truth; synths are generated possibility. Need a believable piano? Sampler — realism is the job, bounded by what got recorded. Need weather instead of an instrument? Synth — no realism obligation, no realism ceiling. Romplers trade depth for speed, which is a real currency. Cast the actor to match the job, then match your programming style to the pretense you picked.
- Good libraries breathe through multisampling, velocity layers, and round-robins — real recordings at every pitch, multiple dynamics per note, several alternate takes per dynamic. Velocity layers are why flat 100s waste an expensive library; round-robins are what kill the machine-gun effect on repeats. And every one of those recordings is triggered by data you write: the realism is rented through your MIDI.
- Program inside the body you're impersonating. Two hands on keys, roughly an octave each, chords that roll like hands falling. One bow, one breath, lines that connect. One bass note at a time, with note length doing the groove work. Realism fails less from missing nuance than from casual impossibility — learn what the instrument can't do and stop doing it. When a part is proudly synthetic, raid these rules as a style menu instead.
- CC data is phrasing's third axis. Velocity is decided once, at the strike; CC moves during the note, which is where sustained music actually lives. CC1 tells the dynamic story, CC11 breathes inside it, CC64 connects piano chords like a foot on a real pedal. The hairpin rule: every held note is going somewhere — swelling in, blooming through, or dying away. Perform the lanes from a wheel when you can; hands wobble in musical ways rulers don't.
- Hands beat mouse, mostly. Anything you record from a controller arrives pre-humanized — your timing leans and your velocities arc because you're a person and it leaks out of you. Half-tempo recording, loop passes, partial quantize afterward: most of the humanization pass, done for free. Mouse-only programming remains completely legitimate; it's just slower at faking what hands give away.
- Sometimes stiff is the point. Kraftwerk, chiptune, house, hyperpop — machine precision as meaning, the grid polished into chrome. Stiff versus human isn't a quality axis; it's an aesthetic axis, and every position on it buys something and costs something. The crime isn't stiffness — it's accidental stiffness. Commit per part, on purpose, in your session notes.
- Finish every part by subtracting. The piano roll makes adding frictionless, which is exactly the trap. Zoom out, play the part in context at your calibrated level, and ask the professional's question: what can leave? Thin the voicing crowding the mud zone, cut the doubled octave nobody hears, mute the clever fill. Less, played better, is the entire recipe.
Remember this
- The notes are the skeleton. The performance is everything else.
- The grid is a reference, not a residence.
- The most expensive orchestra in the world, fed flat 100s, sounds exactly like the doorbell.
- The crime isn't stiffness — it's stiffness nobody chose.
🎚️ "Static Bloom" status
The session finally contains music. The harmonic bed exists as humanized MIDI: an Am–F–C–G pad at 96 BPM — chords rolled bottom-to-top, pad notes starting early so the slow attack blooms on the downbeat, velocities arcing across the four bars, one performed CC1 swell — over a strictly monophonic bass sketch that leans a touch late and leaves daylight before every downbeat for a kick that won't exist until Chapter 13. Both sounds are placeholders, and that's the point: when the real "static bloom" patch gets designed in Chapter 14, every nuance programmed tonight survives the swap untouched, because MIDI is instructions. Saved as static-bloom_v0.3_harmonic-bed.
Pointing forward
Chapter 10 turns to the last untrusted component in your monitoring chain: the room itself. The clap test, the mirror trick, and the $80 treatment pass that outperforms a folder of plugins — because you mix what your room lets you hear, and right now it's editorializing. Bring your humanized harmonic bed; it's about to sound different in a room that's stopped lying.