Sampled, Modeled or Hybrid? Why H.A.T. Brass and Woodwind VST Plugins Sound Real

Sampled, modeled or hybrid: how realistic virtual brass and woodwinds work

Brass and woodwinds are the hardest instruments to reproduce in a VST plugin. A piano note is decided the moment the hammer hits the string; a trumpet or a clarinet note keeps changing for as long as the player blows. There are three ways to build such an instrument: record it (sampling), compute it (physical modeling), or combine the two. Our VSeries instruments – VHorns, VBrass and VWinds – combine the two with H.A.T. (Harmonic Alignment Technology). Here is why.

Sample libraries

A real tone, frozen in place. A sample library is a set of recordings: every note, at a few dynamics, in several articulations (sustain, staccato, marcato, legato...). Play a key and the plugin plays the matching recording.

The tone is real, but a wind note does not stay at one dynamic, and a recording does. To swell from piano to forte, a sample library crossfades between two recordings made at different dynamics. They never line up perfectly, so the sound thickens, blurs or phases in the middle of the swell – you hear two trumpets fading past each other instead of one player blowing harder.

Everything else has to be recorded separately too. Each articulation is its own recording, chosen with keyswitches while you play. Each legato transition you want to hear has to exist as a recording, so the most common ones are covered and the rest are approximated. The result is libraries of tens of gigabytes that still sound best when the part is written around what was recorded.

Physical modeling

Flexible, but synthetic. A physically modeled instrument contains no recordings at all. It computes the sound from a mathematical model of the instrument: the tube, the lips or the reed, the breath pressure, the valves.

Because everything is computed, everything is continuous: dynamics, transitions and bends follow the player's breath, and the plugins are tiny. But the whole tone is synthesized. No model captures every detail of a real instrument and a real player – the grain of the attack, the noise of the air, the small irregularities of a human performance – and the ear notices. Modeled instruments are very expressive, but many listeners can still tell they are not a recording.

H.A.T.

Modeling applied to real recordings: the best of both worlds. H.A.T. does not replace the recording with a model. It applies modeling techniques to the recorded audio itself.

The recording keeps everything that makes it sound natural. Every note you hear is a real player on a real instrument: the natural evolution of the sound from the attack to the release, the player's own tone and breath, and the physics of the recording – the room, the microphones and where they stood. None of that has to be imitated by a model, because none of it was taken out. Each instrument was played by a different world-class musician and recorded with several microphone positions.

The modeling adds the playability. The recordings are harmonically aligned, which lets them be shaped in real time instead of simply played back:

No crossfades. The tone morphs continuously from the softest to the loudest note – one instrument getting brighter as the player blows harder, exactly as on the real thing.
No keyswitches. Staccatos, marcatos, legatos and double tonguing come from how you play the air flow, from the mod wheel, a breath controller, a wind controller or simply note velocity.
Every transition. Legato transitions are modeled from detailed measurements of real ones, from lightning-fast note changes to slow bends, so every interval sounds natural – not only the ones that happened to be recorded.
A living performance. Vibrato moves pitch, volume and timbre together as a player's does, and small variations of air and pitch keep repeated notes from sounding identical.

And it stays small. Because the modeling does the work that a sample library does with thousands of extra recordings, an H.A.T. instrument needs little disk space, little memory and little CPU: a whole section is a few hundred megabytes to around 1.6 GB, and runs comfortably on a laptop.

Compared with a sample library, an H.A.T. instrument plays like the real one. Compared with a purely modeled instrument, it sounds like the real one – because it is the real one, recorded.

At a glance

	Sample library	Physical modeling	H.A.T. (sampling + modeling)
Sound source	Real recordings	Computed	Real recordings
Dynamics	Crossfades between layers	Continuous	Continuous
Articulations	Keyswitches	From your playing	From your playing
Legato transitions	Only the recorded ones	All, synthesized	All, from real recordings
Size	Tens of GB	Very small	Small

What sounds real

What makes a virtual wind instrument sound real:

Continuous dynamics: the timbre gets brighter as the player blows harder, not just louder.
Legato transitions: how one note becomes the next – the fastest give-away of a virtual instrument.
Vibrato: pitch, volume and timbre changing together, not just a pitch wobble.
Small imperfections: no player hits every note perfectly in tune with perfectly steady air.
The room: where the instrument sits, and how the microphones hear it.

Questions

What is the most realistic brass VST plugin?

For a part that has to phrase like a real player, an instrument that combines real recordings with modeling, such as our VHorns and VBrass, keeps both the real tone of the instrument and the continuous dynamics and transitions of a live performance. Listen to the real-versus-virtual comparisons on each product page and judge for yourself.

What is the difference between a VST plugin, a sample library and a virtual instrument?

They are mostly the same thing seen from different sides. A virtual instrument is the instrument you play on your computer; VST (with AU and AAX) is the plugin format your music software loads; a sample library is the set of recordings a sampled or hybrid instrument plays.

Do I need a breath controller?

No. The mod wheel works well, and our instruments can even take the air from note velocity. A breath controller or a wind controller such as an EWI, a Roland Aerophone or a Sylphyo makes phrasing even more natural.

Do these instruments need a powerful computer?

No. Because the modeling does part of the work, they are light on disk space and memory, and run comfortably on a laptop.