Best Piano - PIANOTEQ

[fmr]

Sample piano libraries nowadays seem to follow many different approaches. The first one is to decide the number of layers to record, in order to emulate as precise as possible the different sound registers of an acoustic grand piano, without noticeable transitions. This is what makes the piano a piano - it has to be able to transtion smoothly from the "ppp" to the "fff" with all the "betweens".

Another is to find a way to come up with the pedal action and its influence in the sound. When the sustain pedal is down, the dampers are lifted, and all the strings resonate sympathetically. Having this effects correctly emulated is a challenge, but also a must for a correct emulation of an acoustic piano. A chord played with the pedal ON sounds much fuller than a chord played with the pedal UP. Also, usually in piano performance the chords are attacked with the pedal UP (strings damped) and this is immediatelly pulled DOWN. This will cause the strings to still resonate sympathetically but with less intensity. Also, the notes that are released (to play other notes) keep sounding, naturally decaying.

The room should be emulated, and currenlty seems to be, only by convolution reverb. Some libraries are recorded with "room". Personally, I think this is NOT the right way.

The piano lid is another challenge. It affects the sound of the piano, and is much more than just a filter. Another challenge that sample libray makers have to face.

Then there are the other pedals - the "Una Corda" (Una Corda is not a special piano, but the action of a pedal who slides the keyboard mechanism so that the hammers do not strike all the strings. The sound becomes less strong, and loses some "body" because the piano doesn't resonate so much).

There is also the "sostenuto" pedal, which sustains only the notes that were played when it was pressed, keeping them sounding while all the others are regularly damped.

The piano is so complex and full of subtleties that when the sample library makers took conscience of all these variables, the libraries started to raise in size, until the tenths of GBs they are today, and became very expensive, because they are hard to produce.

Then came the modelled and hybrid pianos, like Pianoteq and True Pianos. True Pianos development, unfortunately, seems to have been abandoned, because the main developer is now working at VSL. Pianoteq has been evolving during the years and is now at version 5. I'd say it still isn't perfect, but reached a state where it competes well with the biggest sample libraries, and has in its favour that it emulates all the subtleties of the a real acoustic piano with much smoother transitions and expressiveness than the sample based pianos, even those with more layers. That's why real pianists seem to favour Pianoteq over the sampled pianos, because, when played in real time, it reacts more like a real piano than the sampled ones, although the best ones still have a big dose of expressiveness.

I think that the future of the virtual piano lays in PM.

The author of the documentary seems to just have Pianoteq. It isn't clear if he tried and compared other virtual pianos, namely the sampled ones. Also, it isn't clear, from the video, which was the real piano and which was the virtual, but the takes at the room sounded noticeably different (namely when they were playing a single note in alternation). Also the recordings heard sounded noticeably different. There are a lot fo things that I would like to see clarified.

[EvilDragon]

Some libraries are recorded with "room". Personally, I think this is the right way.

I disagree, personally, because with that you get unnatural layering of room reverberation on each played note. Not to mention building up the noise floor. Quite unlike real life.

[JonHolstein]

Are you really sure about that?
Are they really playing back a unmanipulated recording of a note? (using sample manipulation, to try to create a sound that is found in a real instrument, is a form of sytnhesis; Sample/synthesis, sometimes called S/S)
Do you know a current sampled piano instrument, that has no manipulation going on in the sustain part?

I don't think what you are saying is a form of synthesis. Anyway, even if it is, I'm quite certain there are piano libraries that do not loop the sustain (Vienna Imperial Grand is the first one that comes to mind). A piano has finite sustain, there isn't a need to loop the sustain the way there is with a synthesizer.

Well, since the sustain is relative to your playing, unless the note isn't triggered until after you played the note and released it, the software can't know how the sustain should behave.

The software could come with a fixed sustain. But I doubt many if any of the high grade sample packs does that.

It could use sample manipulation, to adapt the sustain to your playing. And yes, this is a form of synthesis, it's not subtractive synthesis, but that is not the meaning of the word, not even close, the actual meaning does not come from musical instruments at all.

The software could also use scripting, and fade the initial sample to another sample that fits your playing. Even if sample fading would not be called synthesis normally, when used as a way to recreate the behaviour of a instrument, I would argue that it is a form of synthesis.


It would be possible to use key-switches, and ingore the actual playing of the keys, to have samples behave the way the player wants them to. Then it could be a straight playback of an audio file, yet offer the player the ability to affect the played sound to have the proper expression.
But I haven't heard of such a piano.

[JonHolstein]

I'm not saying samples are useless.
But as far as I know, sample based pianos, do not simply play back audio recordings.
But some people that argue for sample based pianos, use the argument, that a sample based piano is at least the real thing.
As far as I know, there is manipulation going on with the sample playback, so it's actually not a snapshot in time of the real thing.
So that would make any such argument invalid.

That does not invalidate any argument for those that prefer the sound of sampled pianos, compared to physically modelled pianos.
A manipulated note sample, my very well sound better than a physical modelled note, with current technology.

[fmr]

Some libraries are recorded with "room". Personally, I think this is the right way.

I disagree, personally, because with that you get unnatural layering of room reverberation on each played note. Not to mention building up the noise floor. Quite unlike real life.

Sorry, it was a typo. I wanted to say it is NOT the right way. I already corrected my post.

[fmr]

I'm not saying samples are useless.
But as far as I know, sample based pianos, do not simply play back audio recordings.

They do, although they play "many" audio recordings, not just of notes, but also sometimes of sympathetic resonances, several pedalling, and several articulations, and all this is usually controlled by scripts.

But some people that argue for sample based pianos, use the argument, that a sample based piano is at least the real thing.

It's as real as a picture or a movie is the "real thing". It's NOT. It's an image of the real thing.

As far as I know, there is manipulation going on with the sample playback, so it's actually not a snapshot in time of the real thing. So that would make any such argument invalid.

This is true for an hybrid instrument like True Pianos (if you compare the size of each piano library in True Pianos, you will notice these are very small, because most of the sound is done through Physical Modelling. But in sample libraries, scripts usually trigger different recordings, and perhaps some clever filtering, EQing and reverb intensity.

That does not invalidate any argument for those that prefer the sound of sampled pianos, compared to physically modelled pianos.
A manipulated note sample, may very well sound better than a physical modelled note, with current technology.

The best piano libraries sound somewhat convincing, but when the piano is pushed to its boundaries (like in really demanding pieces, like some Beethoven, Chopin, Liszt, Rachmaninoff, and others) we notice that sometimes they lack dynamics, expressiveness and depth. That's where PM wins, in dynamics, expressiveness and depth. This doesn't invalidate that, sometimes, Pianoteq doesn't sound exactly realistic too. But considering the progress done, I'm convincing they will reach a state of realism that will be unachievable by sample based libraries.

[JonHolstein]

I'm not saying samples are useless.
But as far as I know, sample based pianos, do not simply play back audio recordings.

They do, although they play "many" audio recordings, not just of notes, but also sometimes of sympathetic resonances, several pedalling, and several articulations, and all this is usually controlled by scripts.

But the question is, if they even play back a note without any manipulation.
At least the sustain, is as far as I know synthesized in some way or another, by manipulation of the sample, physical modelling, filtering, or perhaps fading between samples, in the higher end sampling packs.

If the articulation is triggered by velocity, pre pressing the pedal, key switch, or by selecting a patch, it would be possible to have an articulation that is a playback of a recording.
But that would limit dynamic playing (no pedaling possible after the note is triggered, and only one note release sustain value, that could be linked to the velocity, but not by actual release of the key).
I dont think high end samplebased pianos work like that. But there could be low end, simpler instruments that do work like that, but then I suspect the actual sample would not be the best one.

[JonHolstein]

As far as I know, there is manipulation going on with the sample playback, so it's actually not a snapshot in time of the real thing. So that would make any such argument invalid.

This is true for an hybrid instrument like True Pianos (if you compare the size of each piano library in True Pianos, you will notice these are very small, because most of the sound is done through Physical Modelling. But in sample libraries, scripts usually trigger different recordings, and perhaps some clever filtering, EQing and reverb intensity.

filtering, EQ-ing, changing reverb parameters, to synthetically recreate a real world phenomena, would be a form of synthesis. I would argue that fading between samples to synthetically recreate a real world phenomen is also a form of synthesis.

The thing I want to highlight is just that even in sample based pianos, you are most likely not hearing an exact playback of a recording.
I've seen people argue against modelling that it's not the real thing, but that a sample is.
I'm just focusing on that argument.

And not which one sounds best to each and everyones ears.

[EvilDragon]

They DO play back the WHOLE note. No manipulation in sustain. If there's any velocity layer crossfading, that is done from the start of the note, not in the middle of it. Same goes for other things, like sympathetic resonance, etc. Everything's done from the start of the note and lasts all the way through the end of the note. No manipulation in-between.

So no, you are not correct here.


And yeah, crossfading different "snapshots" is totally NOT what happens in real life, so no, samples are NOT the real thing, and they never could really be.

[logburner]

They DO play back the WHOLE note. No manipulation in sustain. If there's any velocity layer crossfading, that is done from the start of the note, not in the middle of it. Same goes for other things, like sympathetic resonance, etc. Everything's done from the start of the note and lasts all the way through the end of the note. No manipulation in-between.

So no, you are not correct here.


And yeah, crossfading different "snapshots" is totally NOT what happens in real life, so no, samples are NOT the real thing, and they never could really be.

Hi EvilDragon,

I enjoy reading your in-depth analysis of software pianos and I'm aware that you are a huge fan of Pianoteq.

I understand that the reproduction of piano sound by sample piano library will never be the same as what happens in the physical world hence leads to your statement that "samples are NOT the real thing". At the same time though, even though modelling technology such as Pianoteq could emulate all the fine behaviours of an acoustic piano, I can't see it being the same as "the real thing" because modelling technology would be synthesising the sound of a piano at some stages, be it sustain phase, sympathetic resonance or soundboard vibration.

[EvilDragon]

That is EXACTLY what makes it behave and sound like the real thing, from where I stand. I just find so many problems with sample libraries, no matter how well they might be done. At one point or another it simply doesn't responding to my input like a real piano would, and it is a huge turn-off for me. It could sound like million dollars, it doesn't matter to me.


And this I say as a sample library developer, as well. I know it might sound a bit hypocritical of me, but it's unfortunately how I feel. Some things are great for samples (i.e. drums), others not that much, even though they might produce passable results for majority of public. To some, this doesn't matter, to others, it matters.

[logburner]

That is EXACTLY what makes it behave and sound like the real thing, from where I stand. I just find so many problems with sample libraries, no matter how well they might be done. At one point or another it simply doesn't responding to my input like a real piano would, and it is a huge turn-off for me. It could sound like million dollars, it doesn't matter to me.


And this I say as a sample library developer, as well. I know it might sound a bit hypocritical of me, but it's unfortunately how I feel. Some things are great for samples (i.e. drums), others not that much, even though they might produce passable results for majority of public. To some, this doesn't matter, to others, it matters.

Okay, I think I understand how you feel about software pianos and your expectations.

Some people say that sample piano libraries work fine for slow passages, but fail to reproduce real-life like responses of an acoustic piano in more expressive phrases. This is more so when you expect a software piano to react to a controller keyboard like a real piano would do.

I generally prefer sampled pianos but they have unnatural decay phase when playing staccato, as I hear damper felt vibration before a note completes decays. There are sampled piano libraries with staccato patches, but sample switching on a sequencer is too much hassle. So I'm open to physical modelling approach to virtual piano, but maybe not yet. I'm sure an improvement in Pianoteq will attract even greater number of users though.

[otristan]

Some sampled piano handles staccato pretty well using multiple release samples

Check out the Ravenscroft one
https://www.vilabsaudio.com/Ravenscroft-By-VI-Labs

my 2 cents

[Cinebient]

I tryed the Pianoteq demo for a longer time and played trough all the emulations it comes with but i find some real samples still sound much more alive and "characterful". Even when Pianoteq can create more realistic dynamics and offers all those tiny things a good piano player might notice.
I'm not a good player and so i might be dissqualified to judge. But especially for more non standard pianos i find there is nothing in Pianoteq which could please me (f.e. something like the wing tack piano in Keyscape).
At the end there is no best software piano tool for everyone as there is no best real piano for everyone.

[Ingonator]

Some sampled piano handles staccato pretty well using multiple release samples

Check out the Ravenscroft one
https://www.vilabsaudio.com/Ravenscroft-By-VI-Labs

my 2 cents

FWIW i have bought the VI Labs Ravenscroft 275 yesterday evening (finally after i already wanted to have it for a while...) and just rendered demos for my piano overview/comparison thread. Will post those soon there (UPDATE: The demos are added threre now).

While i have used it only a short time now overall it sounds really impressive indeed.

It also works in UVI Falocn which offers many interesting options.