Why Your Strings Sound Fake, and It Is Not the Library You Bought

You bought the good library. You read the roundups, you spent the money, you loaded the patch, and the mockup still sounds like a mockup. The tone is right and the room is right, and something is still wrong.

It is almost never the samples. It is when the notes start.

A sampled string section gives itself away in the first 100 milliseconds of every phrase, before any question of tone or reverb arises. Fix that and the same library you already own sounds like a different product.

Measure What Your Legato Patch Actually Does

Every sampled note begins before it has a pitch. The engineer started the recording as the bow moved, so the file opens with rosin and pressure and hair, and the note you actually hear arrives some distance into the sample.

That distance is a real number and you can read it in thirty seconds. Bounce a single sustained note from your legato patch to audio, starting exactly on a bar line, and zoom in until you can see the waveform. Measure from the bar line to the point where the waveform stops being noise and starts being a periodic tone.

On lyrical legato patches, expect 60 to 110ms. On sustains without the legato transition, 40 to 70ms. On spiccato and staccato, 5 to 20ms. Those are the numbers your ear has been complaining about.

The annoyance here is that the figure is not constant across the instrument. Low notes on a bass patch speak later than high notes on a violin patch, sometimes by 30ms, because the string physically takes longer to settle. Measure at the register you actually write in rather than at C3 out of habit.

Set Negative Track Delay Per Articulation, Not Per Instrument

This is the single change that does the most, and most people apply it at the wrong level.

The common approach is one track delay per instrument: violins get -60ms, cellos get -60ms, done. That fails the moment the part switches articulation, because the legato you corrected for and the spiccato on the same track need offsets that differ by 80ms. Correct one and you break the other.

So split articulations onto their own tracks, or use a DAW that allows delay per MIDI channel, and set each one against its own measured value.

Treat those as a starting point, not a prescription. They come from measuring a handful of Kontakt orchestral patches, and your library will differ.

The friction is real and worth knowing before you commit. Cubase exposes track delay in milliseconds in the inspector, Logic hides it under the track header's advanced parameters, and Pro Tools handles it differently again through delay compensation. In all of them, negative delay means the DAW must look ahead, so a very large negative value on a track that also carries live input will fight you.

Playing connects you to YouTube, which may set cookies.

Stagger the Section Entries in the Order They Actually Speak

Once every articulation lands where you asked, a second problem becomes audible: the whole section starts at exactly the same instant, which no real section has ever done.

Sixteen players do not begin together. The lighter instruments speak first because there is less string to move, and the spread across a full section is something like 8 to 18ms from first violins to basses. That spread is not random, it is ordered, which is why the humanize button does not fix it.

Set it as a fixed offset per section, on top of the articulation delay:

Reach for randomisation only inside a section, and keep it small, around 4ms either side. Anything larger and a unison line starts to flam.

Here is the mistake I made for two years. I used the DAW humanize function, which randomises note start and velocity together, and on a library where CC1 carries the dynamics the velocity half of that does nothing at all while the timing half applies the same random spread to a solo cello and a full section. The result sounded less mechanical and no more like an orchestra.

Record the Line, the Dynamics and the Note Ends Separately

Three passes, in this order, on every part that matters.

Pass one is notes only. Play or draw the pitches and rhythm with the mod wheel parked around 60, and do not touch it. You are establishing the line, and nothing else.

Pass two is the CC1 ride. Play the phrase again with your hand on the wheel and your ears on the shape, ignoring the keyboard entirely. A fader with a long throw is easier than a wheel for slow swells, and either beats drawing the curve with a mouse, which produces the ramp shape that says mockup more loudly than anything else on this page.

Pass three is note ends, and it is the one everyone skips. A real bow releases a note over 100 to 200ms. A sample cut off at the note-off gets its release layer chopped, so shorten your long note-offs by 20 to 40ms and give the release somewhere to sound.

The catch on that third pass is legato. Sampled legato transitions usually require the notes to overlap, so if you shorten aggressively across a phrase you will break the transition and get a re-attack on every note instead. Shorten the last note of a phrase and the notes before a rest, and leave the interior of a legato line alone.

Test It at Half Speed Before You Trust It

Bad timing hides at tempo and cannot hide at half speed.

Drop the session to 50% and listen to eight bars. Attacks that felt acceptable at 120bpm turn into obvious little clusters of notes arriving at the wrong moment, and the parts that are genuinely wrong separate themselves from the parts you were only worried about. Then take it back to tempo and fix the ones you heard.

The other cheap test is to mute everything except strings and percussion. Percussion samples are near-instant, so any string part still fighting its attack delay will drag audibly against the grid that the percussion is defining. If the strings sit against a marked-up snare, they will sit against anything.

Both tests take under five minutes and neither requires you to trust your memory of what an orchestra sounds like, which after nine hours in the same room is not worth much anyway.

What This Does Not Fix

Timing is the largest single tell, and it is not the only one. Once note starts are correct you will start hearing the things underneath: a CC1 ride that is a straight ramp rather than a shape, block chords where a real orchestra would divide the line, one reverb doing the work of a hall and a room together.

Those are separate problems with separate fixes, and none of them are audible while the timing is wrong, because the ear is busy with the rhythm and never gets to the tone.

Start here. It costs an evening, it applies to every library you already own, and it does not require you to buy anything.