Sampled or Modeled? Better or just different? (Page 1 of 2)

Yes, they're very different. This is not a blind comparison test like the last two files I posted. It's merely to see which sound people like better. I know what I like best, but I thought I would add my comments a little later in the thread, and to the modeled versus sampled controversy.

Last night I spent 4-5 hours listening closely to a well sampled piano and then adjusting various parameters in a PianoTeq preset to try to sound almost identical to the sample. But I became very, very bored doing it; I felt like I was making PianoTeq's sound worse and worse. I have played sampled pianos for years, and more recently, Roland's overhyped (in my view) "SuperNatural" pianos. I like PianoTeq much better.

Both are good, both are too bright for my taste, but I like "A" more. "B" has slightly synthetic character, so I think it's Pianoteq.

Funny fact is that nobody from audience really care about sound. Audience care about music, sound just should be not so bad. And even internal sound of cheap digital pianos most people consider as not so bad.

Thanks skip and everyone for your comments. Overly bright file B removed, replaced with less bright file C.

"A" has a lot more clarity than "C"; this is very apparent in the block chords starting around 0min 22sec.

edit: I've repositioned my speakers and listening position slightly and the difference in "clarity" I mentioned above is no longer so apparent. I believe I was smack bang in the middle of the room (5m x 4m approx) before.

Hi, i hear in A a clear better sound but mis the dynamics, in C I hear a warmer deeper sound and better dynamics but more synthetic sound,
Thanks en good luck!

"D" is very synthetic, my brain doesn't believe that it's a piano.

I prefer the C.
I don't like the D at all

File A is a sampled piano.
File C is a PianoTeq model.
File D is a piano tone that has been highly processed, has no sympathetic string resonance, no delay, no reverb, and a large degree of equalization applied.

To me, most piano samples, however clear and beautiful, sound relatively flat, sterile, and dead.
PianoTeq's models sound dynamic and alive.
These are of course subjective impressions. But I would rather see a giant Blue Morpho butterfly alive in the rainforest (as I have done) than a dead specimen of the butterfly in a display case. I would rather see a real opal or diamond, sparkling from different angles, than the best static photograph of either gemstone.

In this sense, although I'm sure that the PianoTeq developers will continue to refine the algorithms that produce the PianoTeq models, PianoTeq seems like an attempt to recreate the "living, breathing" dynamic sound and environment of an acoustic piano, but more than that, to create one or several ideal versions of those pianos it attempts to model, lacking many of the deficiencies, eccentricities, unwanted and erratic resonances or other tonal characteristics in certain small ranges or even single keys or notes that cause uneven gradations in volume or marked differences in tone color, less than ideal soundboards, poorly manufactured, damaged or flawed strings and who knows what other odd characteristics pianos usually have from one instrument to another. PianoTeq even allows one to add such variation from one key or tone to the next to simulate the variations present in nearly all acuoustic pianos.

An acoustic piano maker might be happy to achieve in a real acoustic piano some of the sound possible or created by PianoTeq, and musicians might enjoy playing such a piano. In this sense, instead of asking, "does PianoTeq sound as good as an acoustic piano," the thought might occur to a person, as it has to me: does any acoustic piano I've ever played sound as good as the PianoTeq models already sound, regardless of how much or in what ways they could be further improved or made more ideal?

Music itself is not composed of the sounds used to realize or embody the music. There are a great many instruments that sound different one from another. We don't complain that a violin doesn't sound like a saxophone, or a french horn doesn't sound like the human voice. Music is composed merely of rhythm, melody and harmony (and sometimes not all three at the same time). Composers and musicians use a great variety of instruments and sounds to produce those rhythms, melodies and harmonies. Even among pianos and other keyboard instruments such as historical fortepianos and harpsichords, the sounds vary greatly, but the music, while sounding different on each, can be as compelling in certain ways using many different types of sound source, which is why so many transcriptions abound, even by the original composer (such as Bach), from keyboard for example to violins and other strings, guitar, human voice or even marimba. A great melody can be recognizable as a great melody on a toy piano or on a concert-hall grand piano. Great rhythms can be played as effectively and movingly on found and natural objects as on well-tuned drums and other percussion instruments.

Since the creation of electronic sound sources, such as the theremin and later, the early Moog synthetic sound modules, musicians and composers have eagerly adopted alternative sound sources to use for the rhythms, melodies and harmonies of the music they compose or play.

To my ears, the PianoTeq models are already usually more engaging and enjoyable than most sampled pianos or synthesizer piano-like tones I've played, even though I continue to use and enjoy some other piano samples and piano-like tones at times, sometimes in conjunction with each other. But for me, the chief goal is to create living music, and PianoTeq seems like a wonderful tool, one among many, to use or exploit for that purpose.

This is just one point of view. I'll step away from the podium and the microphone now.

Steve

This is an amazing statement the like of which I don't think I've heard anyone say before. The Modartt guys will be happy when they read it.

There are nothing bad in various electronic sounds, but if a certain Pianoteq's model promote itself as virtual copy of a certain acoustic piano, it should be sounds as close as possible to that piano. So, as I understand from this topic I actually not the only man who hears the synthetic character of Pianoteq's examples here. It's may be ok for K2 model, but for all models that copy a real acoustic piano it's a bad thing and should be fixed.

And yes, Pianoteq models are more "live" than sampled. But unfortunately only the pianist feel it, because only he can estimate a response. Audience only listen the final rendering sound and maybe will prefer a sampled piano, that actually was showed in this topic.

I'm not a fan of Pianoteq, and not an enemy. I just try be objective to criticize it.

Dick: Thanks - I like A best, but not by much.

Greg.

Great comparison, Dick van Dyke. Both files sounds great.

A is more clear, authentic, and has better attack.

B has better dynamics, sustain and damping, wider timbral range and more dependence on velocity, more consistent, but sometimes sounds too synthetic (for example, at 2:00).

For overall sound I prefer A, but not so much. Maybe because A's advantages are more central and immediately apparent, but B's advantage are more subtle.

I like "C"  the best.  It seems to have more vibrance and clarity.  I'm listening with headphones.  Curious to know which is which.  There's a lot of variation in Pianoteq voices so one piano is not necessarily the best for each musical style.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Ha ha, I chose the Pianoteq simply based on my own personal taste of musicality.

I absolutely prefer my Pianoteq to my 5'10" Chickering Grand.  The bottom line is Pianoteq is a very musical 21st Century Instrument.

@slobajudge --- Interesting comments and point of view. However, regarding the following comment--

A sampled piano is not "a real piano" either, but instead a recorded approximation of an acoustic piano, which usually sounds different live than in a recording. With a sampled piano we are not hearing the piano, we are hearing the microphone.

@slobajudge -- Sampled pianos can certainly sound very good. For years my favorite piano sound was the sampled "Perfect Piano," which like Fluffy Audio's My Piano, was the work of one man, musician and recording engineer William Coakley. It was a beautifully sampled piano sound that I still have and use occasionally, although it sounds a little sterile and flat to me now when I compare it to more dynamic piano representations that include algorithm based characteristics such as modeled sympathetic string resonance (which I see Fluffy Audio uses as well).

We hear recorded piano performances, sampled performances, and modeled piano performances through headsets or speakers.

Strictly speaking I would say we are essentially listening to speakers systems of one sort or another, transmitted to the human ear. Then making a judgement about the instrument sound of the speaker system.

For myself and a few others, we can't tell the difference between Pianoteq or Sampled pianos or recorded live pianos be it Steinway or any other brand heard in this way (barring some obvious giveaway, and even then). Last year or so, on the Pianoworld forum there was a blind taste test between pianos and I happened to choose Pianoteq because I thought it was the best sounding piano. I was not the only one to choose Pianoteq.

I also own the Ravenscroft VST and I just feel that it sounds synthetic through my headset (AKG 240 MK II). I still like it, but it is a real piano sampled sound. Pianoteq D4 sounds exactly like a real Steinway D recorded classical performance that I heard on Youtube sometime ago.

When I have money I would like to see if I could create a 5:1 Speaker System that replicates the feeling of the waves of vibrations when sitting at an acoustic.

@scherbakov.al -- Those are very interesting images and video of a comparison between the spectral and waveform analyses of PianoTeq and an acoustic piano. Thank you for posting those.

And, yours is a beautiful performance of the Scriabin Etude opus 2, number 1 at YouTube (and it sounds great using PianoTeq)--

https://youtu.be/tntJR-YmD8k

Very, very nice!

@scherbakov.al:

Reviewing it quickly, I'm not sure that I understand the video that you posted. Which notes are from Pianoteq and which are from the real piano?

In the images, I do see the difference in the regularity of the beats. But I'm not sure that the regularity of the beats is the only difference:

1. Does the real piano note also have more beats--more beats per second-- in the middle partials, but have fewer beats on the upper partials than Pianoteq?

2. The amplitude of each beat, the degree to which each partial reduces to 0, seems to vary more in the real piano.

3. The low formant partials last longer on the real piano than on Pianoteq. For example, in your first image, the lowest three partials continue to throb in the left-hand image. In the right-hand image, their volume fades much faster. Particularly on the lowest partial.

Not trying to critique your critique. Just couldn't help noticing other differences. In any case, this kind of analysis, comparing notes from original instruments and from the results of the model, is very good. Thanks for posting these images. More spectral images from many people may be still better, assuming that they are from the same make and model of the piano?

And the exact tuning of each string, naturally, may also affect some of the things I listed above in comparing the two notes. However, I don't think it would affect the duration\continued throbbing of the lower partials.

Of course, the ideal situation would be to compare recordings of the actual piano used as the point of departure to the equivalent modeled notes, but only Modartt can do that. And I think I can understand the problem in trying to create a more exact duplication of notes. Some qualities of the notes may arise from small flaws in the instrument that are hard to identify, or a combination of flaws. The degree to which the silences are actually silent, for example, could arise from wear on strings, the metal used in the strings, the  exact tuning, and other hard-to-trace factors. My first thought would be that coding for these would be difficult, since any equation would tend to create regularity--as one variable changed, the resulting effect on the sound would change in direct proportion. Perhaps not, however, if the variables could operate within limits and\or if the correct differential equations could be determined, with some variables carrying more weight than others?

In any case, I do hope that more people will post more comparisons of notes. Regardless of what is found, new information will open up new areas for experimentation and for calibrating the model.

I wouldn't call 10 years of Pianoteq's development exactly "young", at all. 10 years is a huge amount of time, especially in computer world.

Yes increasing the unison detuning of Pianoteq may help. However, I also hear a kind of high pitched metallic resonance in the acoustic recording, which is completely absent in this Pianoteq recording.  As a pop music lover, I like that kind of sound in a piano. ;^) Is it possible to tweak the Pianoteq preset to produce those overtones?

Greg.

Personally i prefer modeled.

Sampled pianos can sound good if they (a) have a sh*tload of samples for every single feature, and (b) use sample offsets, for example you can get pretty good sympathetic resonance by playing low volume note samples past the attack/decay part, with just a bit of lowpass filtering for good measure, and (c) clever -and well applied- filtering tricks for effects like lid angle, dampers touching the vibrating strings, etc etc.

Then again i'm far from being an expert, this is just my personal opinion

Jake Johnson, the energy between the partial - is noise while recording. Noises premises - ventilation, lighting, street noise of the recording device. So as I was a little increased the sound level - increased noise. Therefore, a piano between the partial orange.
If I placed the recorder another example closer to the strings and soundboard, it is possible to get more quantity and large volume of the upper part by the record.