Which Piano VST Should I Buy?

It’s 2019 and there are literally hundreds of piano sample libraries on the market. 

How do you decide what you need—especially if you’re new to the game? 

Here are some things you should know and watch out for before spending money on a virtual piano:

Decide whether you need a sampled or modeled instrument.

Sampled Instruments

A sample is literally an audio recording of an articulation, note, or phrase played by a real instrument. Each “sample” is assigned to a bank in a sample host plugin (e.g. Kontakt, UVI, Halion), which is assigned to a MIDI note (usually the same pitch and register, depending on the instrument). This can be triggered by playing it on your keyboard or within your MIDI editor. 

Basic sample libraries will have a few layers of expression per note (e.g. a piano playing p, mf, and f dynamics). This means that a piano, which has 88 notes will have three velocity layers per note. That equals 264 samples in total. This is not a large sample library and it means you will likely hear the same sample repeated in your music. This takes away from its expression and nuance. Furthermore, many stock piano libraries don’t even record every note on a piano, but rather, use a technique called “resampling” where every third half step or so is recorded and software is used to alter the pitch of the recorded samples to fill in the gaps. This method saves on space, but it means you’re not getting the true sound of some of the notes. 

As you ascend into higher quality sample libraries, you will begin to acquire more features such as: 

• Higher velocity layer count
• As mentioned earlier, 3 layers per note is the standard for achieving some expression in your piano sound. However, as computers are equipped with more storage, memory, and CPU, companies are continuously adding more velocity layers. This means that not only will you have pp, p, mp, mf, f, & ff (six velocity layers) per note, you will also have velocity layers in between to give you a smoother transition between dynamics. 
• Multiple samples per note
• Even though velocity layers are technically the same note, multiple samples are taken of the same note and same dynamic to achieve even more expression. So a C4 key can have 4 different recordings of p, mf, and f, which means there would be 12 samples for one single note as opposed to 3. So an 88-key piano library would have a total of 1,056 samples. Having multiple samples makes use of a technique called “Round-robin” in which different samples of the same note and same velocity are used in order to avoid the “machine gun” effect where the same sample is used in succession—an unrealistic effect. Clap your hands several times. The sonic print of the sound will change slightly every time.  
• Multiple mic positions
• Most piano sample libraries are recorded in stereo to give a sense of space. However, larger, higher-end sample libraries will record samples using different mic positions (player, side, room, close, etc…). This means that the number of samples will be multiplied by the number of available mic positions. This will increase the size of your sample library, but it will also give you more options on the sonic image of your piano. For example, your piano can go from intimately close to sounding like it’s in an orchestra pit far away from you. Or you can blend them to get a spatially rich sound. 
• Sympathetic resonance
• Sympathetic resonance has been around for a long time, so even cheap sample libraries may have it. However, it’s not to undermine how important it is. When you strike a key on the piano, the vibration and sound of the string affects all the other strings, especially those of the same pitch. When the sustain pedal is depressed, the damper will be lifted from the strings and will allow them to vibrate, creating some beautiful overtones and adding complexity to the sound. 
• Release samples
• It’s obvious that when you strike a key, you will hear sound. What isn’t obvious is the noise the piano makes when you let go of the key. A release sample is a recording of the short sound the strings make when you release a key and the damper returns to the strings to mute them. It’s a pleasant sound to hear. Without it, the piano sound seems to cut off quickly. 
• Damper Noise
• If you play a real piano, you will notice that when you depress the sustain pedal, the damper will lift from the strings allowing them to vibrate. This creates a very soft “fuff”sound. It sounds like the piano is taking a breath. Some people enjoy this sound, some don’t. High quality piano libraries will let you adjust the volume of this sound. 
• Una corda
• Una corda is Italian for “one string.” Pianos have 2-3 strings per note to achieve a wide dynamic range. The una corda pedal (or soft pedal) moves the key bed so that the felt hammers strike 1-2 strings instead of 2-3. This allows the piano to play softer and gives it a warmer tone. Higher quality sample libraries will have these samples available. 
• Envelope
• When a piano note is played and held, its sound can last almost 30 seconds! Multiply this by 88 keys, multiple dynamic levels, multiple samples per dynamic levels, and multiple mic positions, and you can see how things can add up really quickly. One technique some companies use to save on space is to “loop” the audio sample at a point where the note is naturally sustaining in order to give the illusion that it is ringing, then artificially create a decay. This saves on memory but does not give a natural sounding decay. High quality libraries will use a full sample. 
• Half Pedaling  
• This is a technique where the sustain pedal is pressed halfway, allowing the low notes to decay somewhat quickly while the higher notes ring out. This feature may be dependent on your host software. It is not commonly used. 

Modeling Based Instruments

Computer algorithms have become so powerful that they can replicate the behaviors of  electronic circuits and acoustic instrument technology. But not only can they replicate it, they can also go beyond what is possible with the real instrument. 

Modeling based pianos give you the flexibility of adjusting the tuning, string length, material, hammer density, size, and condition of your piano—in addition to many of the advanced features mentioned above (check out Pianoteq 6). 

However, modeled instruments do not use samples, but rather a system of equations that feed off your CPU rather than your RAM in order to produce sound. This means it may only take up a couple hundred megabytes of space on your computer, but it will use a lot of your CPU resources. 

The advantage with modeled instruments is that they have much higher levels of expression because they are not dependent on the quantity of samples. They create new ones depending on the force of your touch. Here’s a clear way of seeing this:

Let’s say a piano sample library has four dynamics—p, mp, mf, and f. That means that a velocity of 1-31 will trigger the p dynamic, 32-63 will trigger mp, 64-95 will trigger mf, and 96-127 will trigger f. A modeled piano would theoretically create a dynamic for every velocity level between 0-127. Not only that, it would create a slightly different sound for a note played at the same velocity in order to avoid the machine gun effect. 

Which is right for you?

It depends. 

What is your budget? Do you want the authentic sound of a specific piano (e.g. Yamaha CFX, Steinway D, etc…)? Do you want flexibility? Is RAM an issue? What host software does it use? Only you can answer that question depending on your needs. 

I hope this helps guide you in the right direction.