How to humanize MIDI strings

The short answer: sampled strings read as human when three details land. Vibrato that starts a fraction of a second late, not on the attack. Bow changes that leave the right mark — a small dip at the join, an accent on the down-bow, silence between martelé notes. And a phrase that arcs and relaxes instead of sitting flat. CC1 carries the arc, CC11 the fine shading, and a third lane carries vibrato — and Palpito drafts all three for you.

Why sampled strings sound stiff

Start with the problem that applies to every section, not just strings: when the CC lanes sit still, every note arrives at the same intensity and the same tone colour from attack to release, and the ear hears "machine" no matter how expensive the library is. That's the ground covered in making a MIDI orchestra sound realistic and in what CC11 does. Strings then add their own tells on top.

A real bowed note is never one steady sound. The bow speed changes across the stroke, the vibrato arrives a moment after the attack rather than with it, and the tone dips slightly every time the bow reverses direction. Flat CC throws all of it away. The three sections below are the three places that detail lives.

Vibrato: the onset delay is everything

No player attacks a note and immediately vibrates at full depth. Vibrato begins after the player has established the pitch and settled into the note — and that gap between the attack and the first vibrato movement, the onset delay, is the single factor that most separates a human performance from a machine one.

How long the wait is depends on the music. An intimate solo line hangs on the bare pitch longest before the vibrato blooms; a lyrical section settles in a little sooner; a heroic, powerful passage leans in almost at once. Short notes get none at all — a spiccato or staccato note is over before vibrato would even begin — and the same is true of pizzicato, where the plucked note decays too fast for vibrato to register, and of tremolo, where the technique already supplies the oscillation.

Two subtler rules keep vibrato from betraying the machine. First, across a slur — several notes under one unbroken bow — the onset delay is paid once, on the slur's first note, and the vibrato then rides continuously through the finger changes; it only restarts at a genuine gap, rest, or phrase boundary. Restart it on every note and the lane drops to zero between pitches, which reads as separately attacked notes instead of a connected line. Second, couple the vibrato to the dynamics: less of it when the line is soft, more when it's loud, a little at a note's start and more toward its end.

There is no standard vibrato CC

Here is the part that surprises people. There is no universal MIDI CC for vibrato across string libraries — every major developer put it on a different control, so no single number is right for the whole market. The layout differs library to library:

• Spitfire (BBC Symphony Orchestra, Symphonic, Chamber) — vibrato on CC21, alongside dynamics on CC1 and expression on CC11.
• Cinematic Studio Series (CSS, CS2) — vibrato on CC2, and CC21 is forbidden there (reserved internally); a CC21 lane has to be remapped to CC2 inside Kontakt.
• Orchestral Tools (Berlin Strings) — CC3 in Kontakt, CC4 in the SINE player, and it switches between discrete vibrato styles rather than crossfading — "nothing in between."
• VSL Synchron Strings — a crossfade fader on CC20 by default, reassignable.
• EastWest Hollywood Strings — vibrato on CC1, the mod wheel, uniquely among the majors.

Every one of those is a factory default, and every one is user-remappable — which is exactly the point.

What is shared is the mechanism, not the number. In nearly every section library, vibrato is a crossfade between two recordings — a non-vibrato (senza) layer and a full-vibrato (molto) layer — blended by a continuous fader, not a synthesized wobble and not baked into a single sample. The fader sweeps depth from none to full, but the curve in between is usually nonlinear, so don't assume a value halfway up gives you half the depth. And solo or leader patches, along with Orchestral Tools, often switch between two fixed states instead of crossfading — a single smooth depth curve can't drive those faithfully.

One place the wobble is simply wrong: an open string. Vibrato comes from the left-hand finger rocking on the string, and an open string has no finger to rock — so vibrato CC on an open note is musically false. Real players don't do it, and it's a small giveaway when a mockup does.

Bow changes leave a fingerprint

A bow is not symmetrical. A down-bow (pulled toward the tip) has gravity helping it, so it starts louder and eases off, with a natural little accent at the onset; an up-bow (pushed back toward the frog) has no such help, so it starts weaker and grows toward the end. Players assign down-bows to downbeats and strong beats for exactly that reason.

That asymmetry matters most when every note gets its own bow. In détaché, martelé, and on-string staccato the player alternates down-bow and up-bow note by note, so consecutive notes carry mirrored shapes — not the same envelope stamped over and over. The first note of a phrase, and each downbeat, is a down-bow; the rest alternate. A believable string line flips that per-note shape every other note instead of repeating one.

Different techniques leave different marks. Legato flows as one smooth arc with no dip at the joins — the bow changes are meant to be invisible. Détaché gives each note a small onset "click," which is the bow change itself. Martelé spikes hard and then goes silent between notes; spiccato spikes and drops to zero as the bow leaves the string; tremolo asks for gentle, near-flat CC11, because the technique already supplies the movement and a big arc would just read as a swell.

Even inside a smooth legato line, each note is shaped from the inside: a touch stronger at the onset, easing through the sustain, then lifting to connect to the next — with a small dip at every note-to-note join that glues the line together. That connecting dip is the "raccord," and it's a large part of why a hand-drawn straight line between two notes sounds stiff and pasted-together.

And on a long held note, real sections do something you can't hear on purpose. The players stagger their bow changes across the desk so no single change lands in the open — the line just sustains, and the staggering glosses the individual strokes into one continuous sound. No notation asks for it; the composer writes the duration and the section handles it. No sample library models those re-bows on a sustain, either — they only appear as note-to-note transitions — so on a long note the re-bow is left to your CC work. When it is worth adding, the spacing is measured in seconds, not beats, and it changes per instrument — a cello's bow is not a violin's, so they re-bow on different clocks. (Louder does not simply mean faster, incidentally — that tidy rule got tested and didn't hold.)

Phrasing: give the line somewhere to go

The bow and the vibrato sit on top of the largest gesture of all, which is the phrase. CC1 carries the broad arc — the rise toward where the line is heading and the relaxation at its close — while CC11 handles the finer shading laid over it, the exact division of labour set out in CC1 vs CC11. Tie that arc to the music's tempo and shape rather than sprinkling random movement, and the same amount of motion suddenly reads as intent. It's the same breath-and-expression shaping walked through step by step in the free guide.

Chords breathe, too. A row of block chords held at one flat level sounds dead; real writing swells up into a chord, dips a touch, then settles into the next — chord-to-chord breathing rather than one plateau. And within a chord the voices aren't equal: the bass and the root carry the fullest expression, while an inner voice holding a note that fights the chord's natural resonance — the third of a minor chord, say — wants a little extra to hold its own. CC can't rescue a badly voiced chord, but it can help a well-voiced one bloom.

One last human detail: players prepare. A swell doesn't begin exactly on the downbeat — it starts a hair before, the way a musician leans in to arrive on the beat rather than after it. The beat is a reference the performance shapes around, not a grid it snaps to, which is part of why precise hand-drawn automation can still feel a step stiff.

Where Palpito fits

The honest position first, the same as everywhere else on this site: a composer riding all of this by hand, with the time and the ear, is unbeatable, and nothing here replaces that. What it costs is hours — a vibrato onset on every sustain, a bow-change dip at every join, a phrase arc per section, repeated across the whole string family — and those hours are what Palpito gives back.

Palpito is a small offline app, built by a working media composer, that reads your MIDI string part and drafts the lanes for you: a tempo-referenced CC1 arc, CC11 detail on top, and a vibrato lane with the onset delayed off the attack, suppressed on open strings, and held once across a slur instead of restarting per note. It's a draft you refine, not a black box — deterministic, so the same input always gives the same result. Fully offline, no cloud, no AI, no subscription. The core app is free.

About that vibrato lane: it lands on CC21 because that's Spitfire's layout, and since no standard exists, some default has to be chosen. Palpito doesn't try to guess your library — it emits one clean CC21 lane, and remapping it to CC2, CC20, CC3, or CC4 for another library is a single step you control from a library-named picker. The one deliberate exception is EastWest, whose vibrato rides the same CC1 mod-wheel that already carries the dynamics lane — there is no separate vibrato control to remap to, so it isn't offered as a target.

The honest test is your own ears. There's a blind A/B on the home page — the same string phrase, the same sample library, only the CC differs — so you can hear what vibrato onset, bow shaping, and a breathing phrase actually change before you decide whether any of this is worth your time.