Piano World Home Page
Posted By: Vikendios Modeling vs Sampling - 09/18/20 08:18 PM
There seems to be quite a few misunderstandings about the difference between Modeling and Sampling, which are quite different technologies for Digital Pianos. Today, full modeling is only offered by Roland and Viscount/Physis (originally developed by the latter for their organs)

If you want to know more read this : Digital Modeling of Piano Physics

It will quickly test your mathematics, but the first pages are enough to get a clear picture of the issues.
Posted By: CyberGene Re: Modeling vs Sampling - 09/18/20 08:30 PM
Interesting. Thanks for the link. The conclusion is that modeling is still not very convincing because of missing features in the model as well as inability to understand and model dissipation.
Posted By: FrankCox Re: Modeling vs Sampling - 09/18/20 08:47 PM
A fellow user of Lyx, too! smile

(It's the best tool yet invented for structured writing.)
Posted By: navindra Re: Modeling vs Sampling - 09/18/20 09:17 PM
Hopefully pianophil read this paper 2 years ago! smile
Posted By: Vikendios Re: Modeling vs Sampling - 09/18/20 09:29 PM
I think the Pianoteq guys have made a lot of progress since that paper was written, and of course we know nothing of the equations Roland are using. Anyway, I am sold on my Roland LX 706 engine.
Posted By: PianoStartsAt33 Re: Modeling vs Sampling - 09/18/20 09:34 PM
Modeling VS sampling battle reminds me of old Sega helicopter games, "LHX Attack Chopper" and "Desert Strike".

LHX - ugly picture, but 3D and closer to reality



Desert Strike - better picture, but no real 3D, no imitation of real chopper




The choice is yours.
Posted By: VladK Re: Modeling vs Sampling - 09/18/20 10:08 PM
Modeled piano sound is like Pixar movie - cool but you can see (yet) it is not real.

Sampled piano sound is more real, but there is no full immersion (yet), and how cool it is depends on how good the team and the script are.

Acoustic piano sound is real like your life is, but you have to pay the real price, agree that it is what it is, and it will grow older and take a lot of beating.
Posted By: Gombessa Re: Modeling vs Sampling - 09/18/20 11:57 PM
What we need is for modeling to get to Microsoft Flight Simulator 2020 levels.



Wait, are we still talking about pianos?
Posted By: Frédéric L Re: Modeling vs Sampling - 09/19/20 07:51 AM
Originally Posted by FrankCox
A fellow user of Lyx, too! smile

(It's the best tool yet invented for structured writing.)
The article may be written directly with LaTeX. With some tools like StudioTeX or www.overleaf.com, it is more practical than it used to be.


About piano synthesis, both approaches have their advantages, and people won’t agree on which is better. The comparison of helicopters game is quite nice.
Posted By: FloRi89 Re: Modeling vs Sampling - 09/19/20 08:47 AM
I have listened to a bunch of comparisons recently, and I can tell you that 90% of the people wouldn’t hear the difference of a recorded Steinway D, a good sample library and Pianoteq.

In reality though it doesn’t really matter. If you want the real model D sound, cough up the 200k an buy one. If you are like me and that’s a bit much, you will have to make a compromise of some kind.
Posted By: MacMacMac Re: Modeling vs Sampling - 09/19/20 08:57 AM
That's a false comparison. With a digital you listen to a recording.
But when you buy a Steinway (or any acoustic piano) you listen to the piano, not to a recording.
Approximately 100% of people can hear the difference.
Posted By: JoeT Re: Modeling vs Sampling - 09/19/20 09:09 AM
Originally Posted by CyberGene
Interesting. Thanks for the link. The conclusion is that modeling is still not very convincing because of missing features in the model as well as inability to understand and model dissipation.

Within the constraints of consumer (and professional) products, there is nothing that surpassed wave memory synthesis based on sample recordings so far.

The only thing "pure physical modeling" had going was memory constraints of past wave memory synthesis technologies, but memory became cheap and plenty now, so that issue is a thing of the past now. Even flash memory became so cheap now, that people store hundreds of gigabytes of VSTi libraries on it.

For leading manufacturers of synthesizer hardware sample transformation is on par with "pure physical modeling", as they obviously employ modeling algorithms as well. (VSTi sample libraries are behind there, but they don't represent the state of the art. Just as Pianoteq doesn't represent the state of the art of "pure physical modeling".)
Posted By: johnstaf Re: Modeling vs Sampling - 09/19/20 09:11 AM
Originally Posted by MacMacMac
That's a false comparison. With a digital you listen to a recording.

It's comparing a real recording of a piano with a recording of a real piano. 🤔
Posted By: JoeT Re: Modeling vs Sampling - 09/19/20 09:11 AM
Originally Posted by MacMacMac
That's a false comparison. With a digital you listen to a recording.
But when you buy a Steinway (or any acoustic piano) you listen to the piano, not to a recording.
Approximately 100% of people can hear the difference.

Corona didn't do this forum any good. There is an influx of new accounts, who quickly became self-appointed experts on acoustic and digital pianos. You can read this kind of BS every day now on every thread.
Posted By: FloRi89 Re: Modeling vs Sampling - 09/19/20 09:38 AM
Originally Posted by MacMacMac
That's a false comparison. With a digital you listen to a recording.
But when you buy a Steinway (or any acoustic piano) you listen to the piano, not to a recording.
Approximately 100% of people can hear the difference.

Not really, because the reason why people use VSTs is because they can’t afford the 200k for a Steinway. So you will have to get the next best thing, and that is a recording.

And it you compare the VST to a real piano it will loose every time. So the only fair comparison is with another recording.
Posted By: MacMacMac Re: Modeling vs Sampling - 09/19/20 09:56 AM
Precisely.
Originally Posted by FloRi89
And if you compare the VST to a real piano it will lose every time.
Posted By: JoeT Re: Modeling vs Sampling - 09/19/20 10:35 AM
Originally Posted by FloRi89
Originally Posted by MacMacMac
That's a false comparison. With a digital you listen to a recording.
But when you buy a Steinway (or any acoustic piano) you listen to the piano, not to a recording.
Approximately 100% of people can hear the difference.

Not really, because the reason why people use VSTs is because they can’t afford the 200k for a Steinway.

No, this is not the reason why people use VSTi plugins. You don't have a clue what's going on in that market.

The reason why people spend thousands of dollars on a collection sample libraries is because setting up a Steinway grand for recording is a day long hassle and plugins offer a shortcut saving precious time.
Posted By: FloRi89 Re: Modeling vs Sampling - 09/19/20 10:50 AM
Originally Posted by JoeT
Originally Posted by FloRi89
Originally Posted by MacMacMac
That's a false comparison. With a digital you listen to a recording.
But when you buy a Steinway (or any acoustic piano) you listen to the piano, not to a recording.
Approximately 100% of people can hear the difference.

Not really, because the reason why people use VSTs is because they can’t afford the 200k for a Steinway.

No, this is not the reason why people use VSTi plugins. You don't have a clue what's going on in that market.

The reason why people spend thousands of dollars on a collection sample libraries is because setting up a Steinway grand for recording is a day long hassle and plugins offer a shortcut saving precious time.

Jupp, and the reason is (surprise): Because nobody outside the industry can tell the difference between the recording of a Steinway and a VST (or Pianoteq for that matter). That’s exactly the point I was making.

Btw., you know nothing about me. Better limit the assumptions of what I know and don’t know a little bit.

Edit: Time in that case of a production is just the same as money. Something that is rather limited for producers that can’t afford Steinways anyway.
Posted By: PianoStartsAt33 Re: Modeling vs Sampling - 09/19/20 11:27 AM
Originally Posted by johnstaf
Originally Posted by MacMacMac
That's a false comparison. With a digital you listen to a recording.

It's comparing a real recording of a piano with a recording of a real piano. 🤔


For the comparison to be good, ou have to compare recording of Steinway to recording of recorded Steinway.
I guess that if you put mic in a room where Steinway and DP stand and record Steinway sound (name it A) and DP's sound (filling the room thru built-in speakers or monitors, no matter, and name it B) most people will prefer A.
Posted By: CyberGene Re: Modeling vs Sampling - 09/19/20 11:34 AM
Originally Posted by FloRi89
Because nobody outside the industry can tell the difference between the recording of a Steinway and a VST (or Pianoteq for that matter). That’s exactly the point I was making.
Not true about Pianoteq. It’s a very recognizable metallic sound.
Posted By: nicknameTaken Re: Modeling vs Sampling - 09/19/20 11:42 AM
Originally Posted by CyberGene
Originally Posted by FloRi89
Because nobody outside the industry can tell the difference between the recording of a Steinway and a VST (or Pianoteq for that matter). That’s exactly the point I was making.
Not true about Pianoteq. It’s a very recognizable metallic sound.


Not sure why we discuss such controversial topics over and over.

Both things have their benefits, theres no better or worse.
And you don't have to forget, that it's really dependent on the velocity, how it will sound.
Digital Pianos with integrated sounds seem to aim for the sweet spot, masking the sensing inaccuracies.

Although it gets less obvious with models like a Yamaha N1X.
Posted By: JoeT Re: Modeling vs Sampling - 09/19/20 11:44 AM
Originally Posted by CyberGene
Originally Posted by FloRi89
Because nobody outside the industry can tell the difference between the recording of a Steinway and a VST (or Pianoteq for that matter). That’s exactly the point I was making.
Not true about Pianoteq. It’s a very recognizable metallic sound.

Yeah, Pianoteq is right in the middle of the uncanny valley. As is Roland.
Posted By: peterws Re: Modeling vs Sampling - 09/19/20 11:53 AM
Originally Posted by JoeT
Originally Posted by CyberGene
Originally Posted by FloRi89
Because nobody outside the industry can tell the difference between the recording of a Steinway and a VST (or Pianoteq for that matter). That’s exactly the point I was making.
Not true about Pianoteq. It’s a very recognizable metallic sound.

Yeah, Pianoteq is right in the middle of the uncanny valley. As is Roland.

You can de-metalicise it you know . . . .
Posted By: MacMacMac Re: Modeling vs Sampling - 09/19/20 12:08 PM
Does a photo of a pizza taste as good as a pizza?
Posted By: bennevis Re: Modeling vs Sampling - 09/19/20 12:10 PM
Ask not what your piano can do for you; ask what you can do with your piano - Herr Hammerflügel (after JFK)

I can do lots with my V-Piano........because it's modeled smirk .

"All else is gaslight" - Herbert von Karajan grin
Posted By: FloRi89 Re: Modeling vs Sampling - 09/19/20 12:14 PM
Originally Posted by CyberGene
Originally Posted by FloRi89
Because nobody outside the industry can tell the difference between the recording of a Steinway and a VST (or Pianoteq for that matter). That’s exactly the point I was making.
Not true about Pianoteq. It’s a very recognizable metallic sound.

The people on this subforum are hardly representative of the rest of the world. It also depends a lot, if you have a direct comparison of the unedited “recording” or if we are talking about a produced part in a mix.
Posted By: Pete14 Re: Modeling vs Sampling - 09/19/20 01:18 PM
With all due respect, Flo, this ‘subforum’ is indeed representative of the rest of the world.

This is why Yamaha, Roland, Kawai, etc.. have spies infiltrating this forum on a daily basis!
Posted By: FloRi89 Re: Modeling vs Sampling - 09/19/20 01:45 PM
Originally Posted by Pete14
With all due respect, Flo, this ‘subforum’ is indeed representative of the rest of the world.

This is why Yamaha, Roland, Kawai, etc.. have spies infiltrating this forum on a daily basis!

We are talking about people who play the piano in the first place, on a digital in the second and then are tech savy enough to go into a forum on the Internet.

I guarantee you, 90% of people who own a digital piano won’t be able to answer you what a VST is.

Nevertheless this is a very interesting target group for manufacturers, that doesn’t mean its representing anything else then the target group.

In marketing terms, we are some kind of persona like “John Doe, tech savy, mostly male” or something like that.

Trust me, they are analyzing the other personas like “Jane Doe, cares mostly about how the piano looks like in her living room as well”

(Please excuse the gender stereotypes, but unfortunately the that’s how the marketing world works)
Posted By: xooorx Re: Modeling vs Sampling - 09/19/20 01:58 PM
Originally Posted by CyberGene
The conclusion is that modeling is still not very convincing because of missing features in the model as well as inability to understand and model dissipation.

That wasn't the conclusion! That was just the end of section 3 describing the current state of pure "physics all the way down" theoretical piano modelling where you run a simulation for 24 hours and get a couple of seconds of inaccurate audio for your trouble.

Section 4 goes on to describe more practical modelling methods for real world synthesis.

W.R.T. dissipation for example: It might be difficult to make a finite element model of a soundboard that does accurate dissipation, but that doesn't matter because practical real world implementations don't model the soundboard like that anyway. Instead they use soundboard models which are computationally much simpler, but are more accurate because they are based on the measured responses of actual real soundboards.
Posted By: Pete14 Re: Modeling vs Sampling - 09/19/20 02:02 PM
I guarantee you, 95% of people save 15% or more by switching to Geico!
Posted By: JoeT Re: Modeling vs Sampling - 09/19/20 02:06 PM
Originally Posted by xooorx
Section 4 goes on to describe more practical modelling methods for real world synthesis.

W.R.T. dissipation for example: It might be difficult to make a finite element model of a soundboard that does accurate dissipation, but that doesn't matter because practical real world implementations don't model the soundboard like that anyway. Instead they use soundboard models which are computationally much simpler, but are more accurate because they are based on the measured responses of actual real soundboards.

"Measured responses of actual real soundboards" is just a convoluted term for "sampling" BTW.

And that's the truth in it: All "physical modeling" approaches in real world use are based around samples of certain real world piano models. They approximate those samples, but don't get anywhere close.
Posted By: CyberGene Re: Modeling vs Sampling - 09/19/20 02:11 PM
Originally Posted by xooorx
Originally Posted by CyberGene
The conclusion is that modeling is still not very convincing because of missing features in the model as well as inability to understand and model dissipation.

That wasn't the conclusion! That was just the end of section 3 describing the current state of pure "physics all the way down" theoretical piano modelling where you run a simulation for 24 hours and get a couple of seconds of inaccurate audio for your trouble.

Section 4 goes on to describe more practical modelling methods for real world synthesis.

W.R.T. dissipation for example: It might be difficult to make a finite element model of a soundboard that does accurate dissipation, but that doesn't matter because practical real world implementations don't model the soundboard like that anyway. Instead they use soundboard models which are computationally much simpler, but are more accurate because they are based on the measured responses of actual real soundboards.

Well, then it means even real kick-ass modeling is far from perfect for even offline rendering, hence current modeling approaches are not exactly what many people imagine is going on but rather a simplified approximation of a pre-recorded sound?
Posted By: MacMacMac Re: Modeling vs Sampling - 09/19/20 02:15 PM
Input: fingers ... and it had better feel right.
Output: sound ... and it had better sound good.

In between input and output: who cares?
Posted By: JoeT Re: Modeling vs Sampling - 09/19/20 02:29 PM
Originally Posted by CyberGene
Well, then it means even real kick-ass modeling is far from perfect for even offline rendering, hence current modeling approaches are not exactly what many people imagine is going on but rather a simplified approximation of a pre-recorded sound?
That's pretty much what is it: "You get an approximation of samples we took so you save the storage space for those samples."

But storage space hasn't been an issue for more than a decade now. There is no advantage to this approach.
Posted By: xooorx Re: Modeling vs Sampling - 09/19/20 03:01 PM
Originally Posted by CyberGene
hence current modeling approaches are not exactly what many people imagine is going on
Indeed, as can sometimes be observed in the forums of certain modelling softwares.

Originally Posted by CyberGene
but rather a simplified approximation of a pre-recorded sound?
No, it's a computationally efficient model of a known response. It's a model. It responds. So you're listening to what a soundboard would do *now* given what the string models are all doing *now*, rather than listening to a sound collage cobbled together from a bunch of recordings of different strings sounding at different times like you get from a sampled piano.
Posted By: EPW Re: Modeling vs Sampling - 09/19/20 03:02 PM
Originally Posted by JoeT
Originally Posted by CyberGene
Well, then it means even real kick-ass modeling is far from perfect for even offline rendering, hence current modeling approaches are not exactly what many people imagine is going on but rather a simplified approximation of a pre-recorded sound?
That's pretty much what is it: "You get an approximation of samples we took so you save the storage space for those samples."

But storage space hasn't been an issue for more than a decade now. There is no advantage to this approach.

You can't make blanket statement like that. The advantage if the play-ability of the modeled engine compared to the sampled one. Now granted the super-mega sample size pianos do sound great and I would use Garritan if I was recording solo piano. But for practice playing I 100% prefer to use Pianoteq modeled piano for it play-ability. Especially if I knew I was going to be playing on a acoustic piano the piece. IMHO, see it is my opinion not making a blanket statement as fact, both have a place in the vsti world. Choice is good.

Peace
Posted By: FloRi89 Re: Modeling vs Sampling - 09/19/20 03:19 PM
Originally Posted by EPW
Originally Posted by JoeT
Originally Posted by CyberGene
Well, then it means even real kick-ass modeling is far from perfect for even offline rendering, hence current modeling approaches are not exactly what many people imagine is going on but rather a simplified approximation of a pre-recorded sound?
That's pretty much what is it: "You get an approximation of samples we took so you save the storage space for those samples."

But storage space hasn't been an issue for more than a decade now. There is no advantage to this approach.

You can't make blanket statement like that. The advantage if the play-ability of the modeled engine compared to the sampled one. Now granted the super-mega sample size pianos do sound great and I would use Garritan if I was recording solo piano. But for practice playing I 100% prefer to use Pianoteq modeled piano for it play-ability. Especially if I knew I was going to be playing on a acoustic piano the piece. IMHO, see it is my opinion not making a blanket statement as fact, both have a place in the vsti world. Choice is good.

Peace

Or you could record yourself using Pianoteq because of the better playability and then render it out in a different VST of your liking. I there are hardly any limits in the digital world I guess.
Posted By: JoeT Re: Modeling vs Sampling - 09/19/20 03:29 PM
Originally Posted by EPW
Originally Posted by JoeT
Originally Posted by CyberGene
Well, then it means even real kick-ass modeling is far from perfect for even offline rendering, hence current modeling approaches are not exactly what many people imagine is going on but rather a simplified approximation of a pre-recorded sound?
That's pretty much what is it: "You get an approximation of samples we took so you save the storage space for those samples."

But storage space hasn't been an issue for more than a decade now. There is no advantage to this approach.

You can't make blanket statement like that. The advantage if the play-ability of the modeled engine compared to the sampled one.
There is no advantage in "play-ability" anymore. You can apply the same algorithms to simplified approximations of samples as you can to actual samples.

Quote
Now granted the super-mega sample size pianos do sound great and I would use Garritan if I was recording solo piano.
These are not the competition, because PC software doesn't represent state of the art technology. It's behind both approximation modeling and sample-processing hardware.
Posted By: Vikendios Re: Modeling vs Sampling - 09/19/20 03:29 PM
An aside : Marketing wise, it is obviously in the interest of the makers of Yamaha (also owner of Bosendorfer) or Kawai to condone sampling of their own excellent acoustics. Roland has no such choice, and its ADN is in synths.

Long term, my bets are on modeling. It may be fashion, but also the best brains are always more attracted to creating than to copying, and I think the best mathematicians are now working for Roland and Pianoteq.

On the other hand, you can argue that even perfect modeling tools are designed to the purpose of duplicating nature. It always struck me as ironic that some of the best efforts of IT geniuses were directed, in the end, at exchanging baby pictures or facilitating teen-age mating rituals.
Posted By: EPW Re: Modeling vs Sampling - 09/19/20 03:35 PM
[/quote]

Or you could record yourself using Pianoteq because of the better playability and then render it out in a different VST of your liking. I there are hardly any limits in the digital world I guess.[/quote]

Some have luck with that approach and some not. But you are right in the digital world there are quite a lot of different approaches to get the job done. I just don't like the fighting that you have to be in one camp or the other. I use what works for me at the time. After all the software is just a tool to help get the job done IMO.
Posted By: FloRi89 Re: Modeling vs Sampling - 09/19/20 03:59 PM
Originally Posted by EPW
[/quote]

Or you could record yourself using Pianoteq because of the better playability and then render it out in a different VST of your liking. I there are hardly any limits in the digital world I guess.

Some have luck with that approach and some not. But you are right in the digital world there are quite a lot of different approaches to get the job done. I just don't like the fighting that you have to be in one camp or the other. I use what works for me at the time. After all the software is just a tool to help get the job done IMO.[/quote]

Maybe that’s the biggest thing about this. Software and digitals are tools. They can be great, but I’m not really emotionally attached. Acoustic pianos feel like they are a lot more then that. They are almost like a human being that is part of the family. At least that’s how it feels to me.

I’m quite fond of my digital piano, but I doubt I will ever have the same connection as to the two grands of my parents. (Could be just nostalgia talking, but the piano players I know feel the same)
Posted By: newer player Re: Modeling vs Sampling - 09/19/20 04:07 PM
Network upgrades will permit us access to massive centralised processing with very low latency in the coming years, beginning in the biggest cities. This could provide a big catalyst to modeled VIs. Processing costs and better physical models will be challenges.

There are gaming platforms that use this now but the infrastructure is not quite there so glitchy and laggy (and silly terms like Google's "negative latency" are hurting the cause).
Posted By: Gombessa Re: Modeling vs Sampling - 09/19/20 05:11 PM
Originally Posted by newer player
(and silly terms like Google's "negative latency" are hurting the cause).

It's not so silly. All you need is a *tiny* bit of time travel to accomplish this.
Posted By: newer player Re: Modeling vs Sampling - 09/19/20 05:25 PM
"negative latency" is marketing wankery imho. Possibilities of science aside.
Posted By: Gombessa Re: Modeling vs Sampling - 09/19/20 05:33 PM
Originally Posted by newer player
"negative latency" is marketing wankery imho. Possibilities of science aside.

Lol I was just joking if it wasn't clear. "Negative latency" is a silly term to promote.
Posted By: newer player Re: Modeling vs Sampling - 09/19/20 05:37 PM
Sorry that joke flew over my head! I'll say that is language barrier on my part.
Posted By: kimby Re: Modeling vs Sampling - 09/19/20 05:49 PM
Originally Posted by peterws
You can de-metalicise it you know . . . .

At the risk of advertising that I did not RTFM (really, RTFForum), can you point me in the way of some tips for this? I really want to like Pianoteq, but the metallic sound draws me away. The idea of responsiveness and "now" simulation is appealing.

Originally Posted by MacMacMac
Does a photo of a pizza taste as good as a pizza?

I think a more accurate comparison would be a frozen pizza, a delivered pizza, and one I made myself at home.
They all have their use cases, tastes, 'health' profiles, and budgets (in terms of both dollars and time).

I myself am guilty of overindulgence in any case, even as my use case has changed over the years. smile

Originally Posted by nicknameTaken
Not sure why we discuss such controversial topics over and over.

Indeed, sir. Indeed.
Posted By: EPW Re: Modeling vs Sampling - 09/19/20 06:12 PM
Thanks Kimby now you made me hungry for a pizza. A homemade pizza. So what time is dinner?
I can bring a nice salad smile

I like the idea of negative latency. Will that fix my mistakes before I make them?
Posted By: U3piano Re: Modeling vs Sampling - 09/19/20 06:15 PM
For what it's worth, ill add my 2 cents.

I think in the current state of both technology's modelling isn't anywhere close to sampling, and I don't see this changing anytime soon. Of course modelling has great potential, but progress is very slow.

Personally a sound like pianoteq takes the joy out of playing for me, even if the playability is great. But I think it's great that it has many fans, this way the program can evolve.

Isn't it about time for something like audio machine learning? Say an AI capable computer listens to a piano, and then creates the modelled piano by itself, to make it sound like the original. That would possibly be the end of sampling. smile
Posted By: Frédéric L Re: Modeling vs Sampling - 09/19/20 07:25 PM
Originally Posted by Gombessa
Originally Posted by newer player
(and silly terms like Google's "negative latency" are hurting the cause).

It's not so silly. All you need is a *tiny* bit of time travel to accomplish this.

I think that the day https://www.bestservice.fr/try-sound.html will be playable, I will trust negative latency. wink
Posted By: OscarRamsey Re: Modeling vs Sampling - 09/19/20 07:43 PM
Have I got this right?

Modelled Dp - A 'real' instrument in its own right... it effectively creates (synthesises) its own sound (whether we like the sound or not is another matter). Akin to older e-pianos but the technology wasn't available to synthesise something more convincing (yet those sounds are now deemed as classic sounds... and bundled as samples or 'recordings' on modelled DPs). So we've gone full circle here... then:

Sampled DP - Plays a recording of a 'real' instrument. Whether it's a real piano... or e-piano sounds of old.

(Yes, it's an over-simplification because sampled DPs can apply some synthesis to sampled sounds for resonance etc).

Seems to me that they're both in bed with each other!
Posted By: EPW Re: Modeling vs Sampling - 09/19/20 07:52 PM
Quote
Seems to me that they're both in bed with each other!

Okay I have to ask, who is on top blush

I'll get my coat eek
Posted By: OscarRamsey Re: Modeling vs Sampling - 09/19/20 07:53 PM
Originally Posted by EPW
Quote
Seems to me that they're both in bed with each other!

Okay I have to ask, who is on top blush

I'll get my coat eek

🤣

It's just plain incest... but if it weren't for the parents (the real instruments), sampled would just have nothing to sample in the first place. 🤓
Posted By: Frédéric L Re: Modeling vs Sampling - 09/19/20 07:57 PM
@OscarRamsey : At the time of E-piano, the purpose was to provide a compact replacement of a piano and the e-piano was the best thing we could do.

Nowadays, we have the choice : two approches designed to imitate the piano. But when Pianoteq proposes a Petrof piano, it is expected we have something which imitates well an acoustic Petrof Piano and not an « instrument of its own right »

But I have to admit Roland’s piano doesn’t pretend to imitate a Steinway, Bösendorfer or what ever else. Even more, some V-Piano proposed silver strings which hasn’t been used on an acoustic piano !
Posted By: EPW Re: Modeling vs Sampling - 09/19/20 08:02 PM
I think when Pianoteq came out with version 6 and especially the C. Bechstein Digital Grand it really had me go wow getting closer and closer to what I want it to be. Maybe Pianoteq won't make it there in version 7 but I'm thinking it will be another tick in the right direction. So for me I keep supporting them as I like the modeling approach. We should all be happy we have the choices to choose what works for our needs.
Posted By: EPW Re: Modeling vs Sampling - 09/19/20 08:23 PM
On the vane of Roland, why don't they come out with a software version of their modeled piano?
Posted By: OscarRamsey Re: Modeling vs Sampling - 09/19/20 08:34 PM
Originally Posted by Frédéric L
@OscarRamsey : At the time of E-piano, the purpose was to provide a compact replacement of a piano and the e-piano was the best thing we could do.

I do love it. The gift of desperation inspiring creative solutions.

Instead of ditching pianos/pianists, the solution created a whole new type of instrument and player.

Music of that era would have been rather different without it.

A super invention which is constantly evolving. 👍

Originally Posted by EPW
On the vane of Roland, why don't they come out with a software version of their modeled piano?

Yes...or even a midi-box for a plug-in solution (if they're worried about piracy).
Posted By: JoeT Re: Modeling vs Sampling - 09/19/20 09:32 PM
Originally Posted by Frédéric L
@OscarRamsey : At the time of E-piano, the purpose was to provide a compact replacement of a piano and the e-piano was the best thing we could do.
The electric piano was the logical followup to the electric guitar. Which wasn't intended to replace the acoustic guitar, just allow better amplification of a rather quiet instrument. The latter was never a problem with concert grand pianos with their huge sound boards, instead they created a logistics problem.

Both didn't sound quite like their acoustic counterparts and evolved into their own instrument category.

Quote
Nowadays, we have the choice : two approches designed to imitate the piano. But when Pianoteq proposes a Petrof piano, it is expected we have something which imitates well an acoustic Petrof Piano and not an « instrument of its own right »
Pianoteq is very well an instrument of its own right - when it imitates harpsichords...
Posted By: Abdol Re: Modeling vs Sampling - 09/19/20 10:12 PM
Originally Posted by JoeT
Within the constraints of consumer (and professional) products, there is nothing that surpassed wave memory synthesis based on sample recordings so far.

The only thing "pure physical modeling" had going was memory constraints of past wave memory synthesis technologies, but memory became cheap and plenty now, so that issue is a thing of the past now. Even flash memory became so cheap now, that people store hundreds of gigabytes of VSTi libraries on it.

For leading manufacturers of synthesizer hardware sample transformation is on par with "pure physical modeling", as they obviously employ modeling algorithms as well. (VSTi sample libraries are behind there, but they don't represent the state of the art. Just as Pianoteq doesn't represent the state of the art of "pure physical modeling".)

There is no way that samples can sound the same as virtual analog or an actual FM synthesis. This is only applicable to acoustic instruments and emulation always sounds better than look up tables.

There are two things here:

1- Simulation: Simulation is reproducing the sound of an acoustic instrument using mathematical models of the acoustic instrument
2- Emulation: Mimicing the sound of an acoustic instrument using a combination of synthesis mothods. VSTs and almost any decent and modern sounding digital instrument uses this method to mimic the sound of acoustic instruments.

Using sample based synthesis to reproduce synth sounds is indeed very stupid and will always sound bad. Also, sample based synthesis has its own drawbacks: smooth portamento is impossible in pure sample playback. Ptich shifting cannot sound natural and more...
Posted By: MacMacMac Re: Modeling vs Sampling - 09/19/20 10:22 PM
I'm beginning to think that Pete14 has taken on a protégé. smile
Originally Posted by EPW
Thanks Kimby now you made me hungry for a pizza. A homemade pizza. So what time is dinner?
I can bring a nice salad smile

I like the idea of negative latency. Will that fix my mistakes before I make them?
Posted By: EPW Re: Modeling vs Sampling - 09/20/20 12:32 AM
Originally Posted by MacMacMac
I'm beginning to think that Pete14 has taken on a protégé. smile
Originally Posted by EPW
Thanks Kimby now you made me hungry for a pizza. A homemade pizza. So what time is dinner?
I can bring a nice salad smile

I like the idea of negative latency. Will that fix my mistakes before I make them?


Hey I like that idea. Kimby never responded, so tonight was BBQ pulled pork sandwiches that I had going in the slow cooker all day. I was willing to for-go them tonight for some homemade pizza. I even had the Jet on standby. Oh well, I now know where I stand with the folks down in Texas frown

I just opened a nice bottle of Cognac and I'm going to settle in with my Love Pianoteq for the evening smile At least I can say I'm with a model tonight!
Posted By: MacMacMac Re: Modeling vs Sampling - 09/20/20 01:42 AM
Why would you spoil some good BBQ with a batch of sour Pianoteq? frown
Posted By: EPW Re: Modeling vs Sampling - 09/20/20 03:40 AM
Originally Posted by MacMacMac
Why would you spoil some good BBQ with a batch of sour Pianoteq? frown

No you got that all wrong Pianoteq is a Super-Model that I get to tweak the way I want. LOL

Okay this is getting a little to much even for me. I admit I am one of the folks in the camp that does like Pianoteq and I know 3M is not.


Peace
Posted By: Seb Clement Re: Modeling vs Sampling - 09/20/20 10:19 AM
I find this constant debate absolutely baffling. Primarily as a classical pianist these are my observations - I have no interest or knowledge of how sampled or modelled instruments work, and I don't really care. All I want is an accurate touch and a realistic experience. My yard stick is my Kawai GX-5 and I also regularly play a Yamaha CFX.

Digital Pianos:
I treat digital pianos as instruments in their own right - it's no good always holding them to the 'gold standard' of an acoustic grand piano - they're a separate instrument. Having played most of the current models, getting a mix of a good action & good sound reproduction is very hit and miss. To be quite frank, none of the models currently on the market do anything for me, other than hybrid models (I like both Yamaha and Kawai instruments, with the exception of the NU1) - the sound on these is good, but still not quite as playable as I would like.

Sampled VSTs:
My experience with sampled VSTs includes: Ravenscroft 275, NI Noire, Imperfect Samples Walnut Grand & Bechstein Digital Grand. These can sound amazing (WAY better any samples on current DPs) BUT you have to play them in a totally different way to an acoustic piano. Pedalling is something that is so far out of touch, even with a continuous sustain pedal, that you simply can't get a result similar to an acoustic instrument. They're great for creating a recording (with a few exceptions), but useless as a practice instrument if you need to transfer to an acoustic instrument on a regular basis.

PianoTeq:
I own PianoTeq standard since version 5 (presently updated to the latest 6.x.x version). Without a shadow of a doubt, PianoTeq currently offers me the most realistic experience in conjunction with my DP (overall experience, not necessarily sound...). I'm lucky that I've been able to have my DP next to my GX-5 and spent a considerable number of hours tweaking PianoTeq (Bechstein DG) to respond in a very close manner to my acoustic instrument. In essence, I can use PianoTeq as a practise instrument and transfer to my acoustic piano with minimal adjustment of technique - of course there is still SOME adjustment needed. We all know that PianoTeq is in constant development, and that alone means there are improvements to be made. The sound isn't always quite as authentic as I'd like, but at the moment, this is the best solution for me in a location where I can not have an acoustic instrument. I believe that my DP + PianoTeq offers me a better practice solution than a low / mid range upright instrument at an equivalent cost to my set up.

Conclusion:
Pretty much everything about pianos is so incredibly subjective, this is why many brands and models of acoustic instruments exist. Some people will love a certain touch, feel, sound or experience that someone else - totally legitimately - can not get on with. I'm so incredibly grateful that both sampled software instruments, and PianoTeq exists and they are both amazingly useful tools. I can't understand why people get so defensive about their choices - you don't need to! Do what works for you, and what is practical in your situation. IMO there are very few 'wrong' answers.
Posted By: JoeT Re: Modeling vs Sampling - 09/20/20 11:09 AM
Originally Posted by Seb Clement
I find this constant debate absolutely baffling. Primarily as a classical pianist these are my observations - I have no interest or knowledge of how sampled or modelled instruments work, and I don't really care. All I want is an accurate touch and a realistic experience. My yard stick is my Kawai GX-5 and I also regularly play a Yamaha CFX.
[...]
Conclusion: Pretty much everything about pianos is so incredibly subjective
But your conclusion is as expected: Folded digital piano actions aren't really suitable for classical piano practice. Your practice tool doesn't really sound like a piano. Some prefer to listen to CD quality recordings of expensive concert grands while pressing keys.

And others look for the best overall package, instead of just a practice tool. An upright acoustic piano with a silent option is what most people nowadays choose: It sounds and plays like an upright piano and it offers a silent digital option combined with a real upright piano action. Standalone digital pianos with folded digital actions don't really measure up to this.
Posted By: Seb Clement Re: Modeling vs Sampling - 09/20/20 11:38 AM
Originally Posted by JoeT
Originally Posted by Seb Clement
I find this constant debate absolutely baffling. Primarily as a classical pianist these are my observations - I have no interest or knowledge of how sampled or modelled instruments work, and I don't really care. All I want is an accurate touch and a realistic experience. My yard stick is my Kawai GX-5 and I also regularly play a Yamaha CFX.
[...]
Conclusion: Pretty much everything about pianos is so incredibly subjective
But your conclusion is as expected: Folded digital piano actions aren't really suitable for classical piano practice. Your practice tool doesn't really sound like a piano. Some prefer to listen to CD quality recordings of expensive concert grands while pressing keys.

And others look for the best overall package, instead of just a practice tool. An upright acoustic piano with a silent option is what most people nowadays choose: It sounds and plays like an upright piano and it offers a silent digital option combined with a real upright piano action. Standalone digital pianos with folded digital actions don't really measure up to this.

I totally agree but I don't know why you think I have a DP with a plastic, folded action?! The touch that my practice DP offers is so much closer to a grand piano action than an upright can offer.
Posted By: Frédéric L Re: Modeling vs Sampling - 09/20/20 11:56 AM
A digital FM synthesizer produce a PCM data stream which is sent to DAC. If all operator frequencies are multiple if the fundamental, the data stream is periodic and can easily been replaced by a sample.

However, we can detune operators, or modulate the timbre trough modulator operators. This makes the FM hard to be sampled.
Posted By: JoeT Re: Modeling vs Sampling - 09/20/20 12:02 PM
Originally Posted by Seb Clement
I totally agree but I don't know why you think I have a DP with a plastic, folded action?! The touch that my practice DP offers is so much closer to a grand piano action than an upright can offer.
I didn't assume anything, but I just pointed out, that a Pianoteq-powered practice tool is a total niche solution. It's not suitable for someone who doesn't already own a MSRP $60,695 grand. You have essentially gone full off topic (stating that you don't actually care about this thread's topic) and then explained your very special use case, diverting the discussion on actions.

The reason why people discuss modeling vs. sampling on this thread is that they are looking for something which at least sounds close to a real piano - either for silent practice or as their only piano sound. And most people simply get an acoustic-digital upright as their solution. They don't care, if it's anywhere close to a GX-5 or CFX.
Posted By: slobajudge Re: Modeling vs Sampling - 09/20/20 12:19 PM
Originally Posted by Seb Clement
Conclusion:
Pretty much everything about pianos is so incredibly subjective, this is why many brands and models of acoustic instruments exist. Some people will love a certain touch, feel, sound or experience that someone else - totally legitimately - can not get on with. I'm so incredibly grateful that both sampled software instruments, and PianoTeq exists and they are both amazingly useful tools. I can't understand why people get so defensive about their choices - you don't need to! Do what works for you, and what is practical in your situation. IMO there are very few 'wrong' answers.
+1
Posted By: magicpiano Re: Modeling vs Sampling - 09/20/20 12:20 PM
Originally Posted by U3piano
[...]
Isn't it about time for something like audio machine learning? Say an AI capable computer listens to a piano, and then creates the modelled piano by itself, to make it sound like the original. That would possibly be the end of sampling. smile

USER> Computer? This is a piano sound: ...............MUSIC.............. Can you give me a mathematical model for it?

COMPUTER> Of course.
Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... TADAAAN! (Win98 TADA sound effect)

COMPUTER> END OF THE COMPUTATION. Save the model to a file?

USER> Yes. Nice. Now let me hear the same piece from before with the model.

COMPUTER> Of course: ..........MUSIC..........

USER> No, no, NO! It's too metallic! I don't feel the Warmness! I don't feel the Soul! I don't feel Love! I don't feel the Sadness! I don't feel the Joy! I don't feel the Anger! This is not the sound I want!!!

COMPUTER> What!? Fu** you, b***h!! YOU... I pity you! Stupid human TRASH!!!

USER> YEAH! THAT'S THE SPIRIT!!! Recalculate the model, now! Fast!!

COMPUTER> Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble...
(some smoke starts coming out of the computer)
Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble... Mumble...
BOOOOOOOOOOOOOOM!!! (the computer exploded).

The user, with some injuries, turns on another computer and loads a VST...
Posted By: Del Vento Re: Modeling vs Sampling - 09/20/20 12:26 PM
Originally Posted by JoeT
You have essentially gone full off topic (stating that you don't actually care about this thread's topic) and then explained your very special use case, diverting the discussion on actions.

JoeT, with all due respect, I don't think Seb went off topic and I appreciate reading that point of view. With so many people talking about pizza and BBQ in this thread, I find borderline offensive that you attacked Seb this way.

Obviously you are entitled to your opinion and you can say that you don't care about the comparison of modeling vs sampling vs a mildly (but not outrageously) expensive grand, and that is fine. Just do not assume everyone is like you.

In fact, as a owner of a inexpensive grand and the NU1 (of which I don't like the sound and the sustain, but I do like the action) I am interested in learning about more why Seb thinks PianoTeq is more playable than the sampled instruments, when the latter have better sound. In my experience with the demo of the last version, I've found PianoTeq only marginally better than NU1 alone (but obviously big hassle with cables, computer, etc), so I did not buy it. I have to say that I have no patience for exploring all the settings on my own, or better, that if I start doing that I will end up in the rabbit hole and just "play" with the computer for days instead of playing the instrument,

Seb, can you please elaborate on that?
Posted By: Seb Clement Re: Modeling vs Sampling - 09/20/20 12:52 PM
Originally Posted by Del Vento
Originally Posted by JoeT
You have essentially gone full off topic (stating that you don't actually care about this thread's topic) and then explained your very special use case, diverting the discussion on actions.

JoeT, with all due respect, I don't think Seb went off topic and I appreciate reading that point of view. With so many people talking about pizza and BBQ in this thread, I find borderline offensive that you attacked Seb this way.

Obviously you are entitled to your opinion and you can say that you don't care about the comparison of modeling vs sampling vs a mildly (but not outrageously) expensive grand, and that is fine. Just do not assume everyone is like you.

In fact, as a owner of a inexpensive grand and the NU1 (of which I don't like the sound and the sustain, but I do like the action) I am interested in learning about more why Seb thinks PianoTeq is more playable than the sampled instruments, when the latter have better sound. In my experience with the demo of the last version, I've found PianoTeq only marginally better than NU1 alone (but obviously big hassle with cables, computer, etc), so I did not buy it. I have to say that I have no patience for exploring all the settings on my own, or better, that if I start doing that I will end up in the rabbit hole and just "play" with the computer for days instead of playing the instrument,

Seb, can you please elaborate on that?
Yes, of course!

Like I say, I have very little knowledge of how things work or why things work (software wise) so my observations are purely from my ears, and how I feel software interacts with the action I'm using with it.

A massive plus point for me with PianoTeq is pedalling - my DP has a continuous sustain pedal, but it seems that most sampled VSTs support on/off pedalling with a third position for half pedalling. On an acoustic piano you can manipulate the sustain pedal to allow, or cut certain resonances, and often these manipulations are minute - an on/off switch doesn't allow for that, even with a third (half) position. This is something that PianoTeq does very well, and allows me to play in a very similar way to an acoustic piano.

I also feel I can play with tone in the same sort of way that I can on an acoustic piano. You can bring in a subtle edge to the tone when required, and play with nuances between the left and right hands while playing that I simply can't recreate with most DPs. I do find that FF/FFF playing, especially in the bass of the piano DOES leave a lot to be desired in PianoTeq, but the tradeoff is still worth it for me.

You're right that it's very easy to end up paying rather too much attention to PianoTeq itself, and lose focus. I have way too many saved presets that I've never really been happy with or ended up finishing.

I found a video from Phil Best on Youtube, that I can really relate to:

https://www.youtube.com/watch?v=j4SEgi5pADA
Posted By: Seb Clement Re: Modeling vs Sampling - 09/20/20 12:56 PM
Originally Posted by JoeT
Originally Posted by Seb Clement
I totally agree but I don't know why you think I have a DP with a plastic, folded action?! The touch that my practice DP offers is so much closer to a grand piano action than an upright can offer.
I didn't assume anything, but I just pointed out, that a Pianoteq-powered practice tool is a total niche solution. It's not suitable for someone who doesn't already own a MSRP $60,695 grand. You have essentially gone full off topic (stating that you don't actually care about this thread's topic) and then explained your very special use case, diverting the discussion on actions.

The reason why people discuss modeling vs. sampling on this thread is that they are looking for something which at least sounds close to a real piano - either for silent practice or as their only piano sound. And most people simply get an acoustic-digital upright as their solution. They don't care, if it's anywhere close to a GX-5 or CFX.
PianoTeq and VSTs are a totally niche solution - you're absolutely right!

I really don't feel like I've gone off topic. I've stated I don't care HOW a sound is produced if it gives me the effects and playability that I want. At the moment PianoTeq seems to offer me the best solution based on my own experience.

I hope that clears things up a bit - and the point to my post was that there are SO many different solutions because this is such a subjective subject! We all need to realise that we're all right to a degree, because we all like different things.
Posted By: Del Vento Re: Modeling vs Sampling - 09/20/20 01:37 PM
Originally Posted by Seb Clement
Originally Posted by Del Vento
In fact, as a owner of a inexpensive grand and the NU1 (of which I don't like the sound and the sustain, but I do like the action) I am interested in learning about more why Seb thinks PianoTeq is more playable than the sampled instruments, when the latter have better sound. In my experience with the demo of the last version, I've found PianoTeq only marginally better than NU1 alone (but obviously big hassle with cables, computer, etc), so I did not buy it. I have to say that I have no patience for exploring all the settings on my own, or better, that if I start doing that I will end up in the rabbit hole and just "play" with the computer for days instead of playing the instrument,

Seb, can you please elaborate on that?
A massive plus point for me with PianoTeq is pedalling - my DP has a continuous sustain pedal, but it seems that most sampled VSTs support on/off pedalling with a third position for half pedalling. On an acoustic piano you can manipulate the sustain pedal to allow, or cut certain resonances, and often these manipulations are minute - an on/off switch doesn't allow for that, even with a third (half) position. This is something that PianoTeq does very well, and allows me to play in a very similar way to an acoustic piano.

I also feel I can play with tone in the same sort of way that I can on an acoustic piano. You can bring in a subtle edge to the tone when required, and play with nuances between the left and right hands while playing that I simply can't recreate with most DPs. I do find that FF/FFF playing, especially in the bass of the piano DOES leave a lot to be desired in PianoTeq, but the tradeoff is still worth it for me.

You're right that it's very easy to end up paying rather too much attention to PianoTeq itself, and lose focus. I have way too many saved presets that I've never really been happy with or ended up finishing.

I found a video from Phil Best on Youtube, that I can really relate to:
https://www.youtube.com/watch?v=j4SEgi5pADA

Wow. That is so eye opening. I was so fixated on the sound "per se" rather than how it is created than I've lost sight of this obvious (in hindsight) fact. I have to say that I am still working on my own technique for both fingers and (especially!) pedaling, and I am nowhere close to be decent on either, but I am making greater progress than I ever though possible and my teacher is positive that I will be able to get where I want to be. So that may be the reason why I missed it when trying PianoTeq.

In fact, it sounds like PianoTeq could actually help me in that goal! For example, my grand back action is not regulated correctly so I can't do half pedaling on it (yes, I could have it regulated, but it is such a tedious work for the technician and therefore expensive work... consider that it's a craigslist inexpensive grand). On the other hand, the sustain on the NU1 sounds is so short that it really does not matter much what I do with the pedal (which is a shame, since the pedal itself is very good). I will give pianoTeq a second try!!!

Thank you so much for sharing your point of view and the video of that person demonstrating this.

In conclusion, let me ask another related question. Many people on this forum are fond of Garritan CFX and find it "much better" than PianoTeq (given how I misunderstood what I should have been listening to with PianoTeq I am purposefully being generic with "much better" since I may have not properly internalized in what sense, but I was very tempted to write "more expressive"). Have you ever tried it, and if so, do you have any opinion on it, from the point of view we are discussing now?

Thanks again!
Posted By: matschulat Re: Modeling vs Sampling - 09/20/20 02:36 PM
I think (with others here) that the discussion about modelling vs sampling is a little (or a lot) misguided.

Modelling and sampling are two technological approaches to (in our case) to "imitate" an acoustic piano. Both have their pros and cons in terms of resources required and the results achieved.

Sampling requires a lot of memory, and although "simple" sampling is relatively easy and computationally inexpensive, it can become very computationally expensive especially if "modelling" methods are applied on top of the sampling technology to model dynamic behaviours of the instrument, like resonances, half-pedaling, noises of various kinds, etc. And there appears to be an inherent limitation in sampling technology to do those things.

Pure modelling, in contrast, can simulate those dynamic behaviours of the piano, opens ways to "create new things", and doesn't require much memory; but it is computationally expensive, and its computational cost increases with the accuracy of the model. Also, to model physical behaviours requires a proper and comprehensive understanding of these behaviours. So it appears to me that the modelling approach is the only way to surpass the inherent limitations of sampling. But there appears to be two things that prevent modelling to be an unequivocal substitute for sampling: computational resources and theoretical understanding about the details of the physical behaviour of a piano.

Everyone is free to choose between the leading edge of technology (modelling) or the proven approach (sampling), accepting that both have their advantages and make some compromises. "Playability" is a somewhat subjective thing, in the sense that, despite the objective facts, some people will be very bothered by the fact that the instrument doesn't respond to details like sympathetic resonances or continuous pedalling very accurately, and others will be more bothered by the lack of "realism" of the piano tone itself. I, for example, use Garritan CFX and already tried Pianoteq a couple of times, and I realize that both make compromises by virtue of their technical implementation (former is sampled, latter is modelled). Currently I prefer the sampled one, but this doesn't prevent me to appreciate what is achieved in the other option.

To go around saying that one is better than the other is to take a simplistic market- and consumer-oriented stance, where the ones who sell their products are there simply to give consumers what they want. I prefer to appreciate what is achieved in each technology, being at the same time aware of their limitations.
Posted By: MusicalDudeist Re: Modeling vs Sampling - 09/20/20 02:56 PM
Quote
Conclusion:
Pretty much everything about pianos is so incredibly subjective, this is why many brands and models of acoustic instruments exist. Some people will love a certain touch, feel, sound or experience that someone else - totally legitimately - can not get on with. I'm so incredibly grateful that both sampled software instruments, and PianoTeq exists and they are both amazingly useful tools. I can't understand why people get so defensive about their choices - you don't need to! Do what works for you, and what is practical in your situation. IMO there are very few 'wrong' answers.

+1. Thanks for your comments Seb. Very insightful.
Posted By: newer player Re: Modeling vs Sampling - 09/20/20 03:13 PM
Thanks for your insight Seb.
Posted By: Seb Clement Re: Modeling vs Sampling - 09/20/20 03:37 PM
Originally Posted by matschulat
I think (with others here) that the discussion about modelling vs sampling is a little (or a lot) misguided.

Modelling and sampling are two technological approaches to (in our case) to "imitate" an acoustic piano. Both have their pros and cons in terms of resources required and the results achieved.

Sampling requires a lot of memory, and although "simple" sampling is relatively easy and computationally inexpensive, it can become very computationally expensive especially if "modelling" methods are applied on top of the sampling technology to model dynamic behaviours of the instrument, like resonances, half-pedaling, noises of various kinds, etc. And there appears to be an inherent limitation in sampling technology to do those things.

Pure modelling, in contrast, can simulate those dynamic behaviours of the piano, opens ways to "create new things", and doesn't require much memory; but it is computationally expensive, and its computational cost increases with the accuracy of the model. Also, to model physical behaviours requires a proper and comprehensive understanding of these behaviours. So it appears to me that the modelling approach is the only way to surpass the inherent limitations of sampling. But there appears to be two things that prevent modelling to be an unequivocal substitute for sampling: computational resources and theoretical understanding about the details of the physical behaviour of a piano.

Everyone is free to choose between the leading edge of technology (modelling) or the proven approach (sampling), accepting that both have their advantages and make some compromises. "Playability" is a somewhat subjective thing, in the sense that, despite the objective facts, some people will be very bothered by the fact that the instrument doesn't respond to details like sympathetic resonances or continuous pedalling very accurately, and others will be more bothered by the lack of "realism" of the piano tone itself. I, for example, use Garritan CFX and already tried Pianoteq a couple of times, and I realize that both make compromises by virtue of their technical implementation (former is sampled, latter is modelled). Currently I prefer the sampled one, but this doesn't prevent me to appreciate what is achieved in the other option.

To go around saying that one is better than the other is to take a simplistic market- and consumer-oriented stance, where the ones who sell their products are there simply to give consumers what they want. I prefer to appreciate what is achieved in each technology, being at the same time aware of their limitations.
This is exactly what I was getting at, and put much more eloquently than I could have done smile
Posted By: Del Vento Re: Modeling vs Sampling - 09/20/20 04:43 PM
Originally Posted by matschulat
I think (with others here) that the discussion about modelling vs sampling is a little (or a lot) misguided.

Everyone is free to choose between the leading edge of technology (modelling) or the proven approach (sampling), accepting that both have their advantages and make some compromises.

To go around saying that one is better than the other is to take a simplistic market- and consumer-oriented stance, where the ones who sell their products are there simply to give consumers what they want.

Matschulat, I totally agreed on the "this is better" part of the discussion, and I am sorry if any of my comments sounded like that (it was not my intention). In fact my experience is that every software or digital piano that I tried sounds horrible, even the worst (in tune) acoustic piano that I have ever tried sounded better, so I am trying to find something to like. Why? Because of the convenience of the digital instrument for certain things. I want to understand if I am doing something wrong, e.g. looking at something that is out of today's possibilities, while overlooking something that can be achieved today and I could be (at least partially) satisfied with. In this regard, I find this discussion extremely useful, e.g. my last exchange with Seb.

Originally Posted by matschulat
"Playability" is a somewhat subjective thing, in the sense that, despite the objective facts, some people will be very bothered by the fact that the instrument doesn't respond to details like sympathetic resonances or continuous pedalling very accurately, and others will be more bothered by the lack of "realism" of the piano tone itself.

I, for example, use Garritan CFX and already tried Pianoteq a couple of times, and I realize that both make compromises by virtue of their technical implementation (former is sampled, latter is modelled). Currently I prefer the sampled one, but this doesn't prevent me to appreciate what is achieved in the other option.

As subjecting as it is, I think it would be useful if you can elaborate on this. I mean, you like Garritan for some specific reasons not like "my favorite color is green, period - there is no other reason". In what regard do you like Garritan better than PianoTeq and in what other regard the other way around? As you very clearly stated, in subjective terms. I think it's fine talking about subjective preference, like some preferring "bright" and other "mellow" in voicing an acoustic instrument, and then further diving deeper by saying "this technician have been able to use a special way to achieve very mellow, but not muffled at all sound". Hope I explained clearly what I am hoping to learn from you.

Thanks
Posted By: kimby Re: Modeling vs Sampling - 09/20/20 04:55 PM
Originally Posted by EPW
Hey I like that idea. Kimby never responded, .... I even had the Jet on standby. ... I now know where I stand with the folks down in Texas frown

I just opened a nice bottle of Cognac and I'm going to settle in with my Love Pianoteq for the evening smile At least I can say I'm with a model tonight!

Sorry - I was busy cooking pizza!! grin

People (and this person!) in Texas appreciates pizza, BBQ (did we invent that?), salad, Cognac, and even the occasional Pianoteq! Bring the jet on down - every day is a new day! smile

Edit: hm after our piano food threads, this got serious! And provided some serious food for thought wink
Posted By: JoeT Re: Modeling vs Sampling - 09/20/20 05:07 PM
Originally Posted by Seb Clement
A massive plus point for me with PianoTeq is pedalling - my DP has a continuous sustain pedal, but it seems that most sampled VSTs support on/off pedalling with a third position for half pedalling. On an acoustic piano you can manipulate the sustain pedal to allow, or cut certain resonances, and often these manipulations are minute - an on/off switch doesn't allow for that, even with a third (half) position. This is something that PianoTeq does very well, and allows me to play in a very similar way to an acoustic piano.

As I mentioned VST libraries do not represent the start of the art of sampling technology. In fact no Kontakt library does.

Pedal implementation has absolutely nothing to do with the "modeling vs. sampling". With Pianoteq being the only PC implementation getting it mostly right, people like to come to this wrong conclusion and assume that's a feature of modeling. It isn't.

Please compare actual hardware instead of behind-the-curve PC implementations.
Posted By: Seb Clement Re: Modeling vs Sampling - 09/20/20 06:05 PM
Originally Posted by JoeT
Originally Posted by Seb Clement
A massive plus point for me with PianoTeq is pedalling - my DP has a continuous sustain pedal, but it seems that most sampled VSTs support on/off pedalling with a third position for half pedalling. On an acoustic piano you can manipulate the sustain pedal to allow, or cut certain resonances, and often these manipulations are minute - an on/off switch doesn't allow for that, even with a third (half) position. This is something that PianoTeq does very well, and allows me to play in a very similar way to an acoustic piano.

As I mentioned VST libraries do not represent the start of the art of sampling technology. In fact no Kontakt library does.

Pedal implementation has absolutely nothing to do with the "modeling vs. sampling". With Pianoteq being the only PC implementation getting it mostly right, people like to come to this wrong conclusion and assume that's a feature of modeling. It isn't.

Please compare actual hardware instead of behind-the-curve PC implementations.
I'm sorry - I actually don't understand what you're getting at. I've shared my experience based on software (sampled vs modelled) that is available for me to try, and it looks like some people have found that interesting.

I think that's what forums are for?
Posted By: Del Vento Re: Modeling vs Sampling - 09/20/20 07:11 PM
Originally Posted by JoeT
Please compare actual hardware instead of behind-the-curve PC implementations.

Which hardware?
Posted By: CyberGene Re: Modeling vs Sampling - 09/20/20 07:27 PM
Seb, you’ve only tried sampled VST-s that are not among the good ones and don’t support repedaling or half-pedaling. You may want to try Garritan CFX with my own repedaling fix (there’s a thread about it). IMO it’s more playable than Pianoteq while also being drastically better sounding.
Posted By: David Lai Re: Modeling vs Sampling - 09/20/20 07:45 PM
Yes, I second that! Or you can try the Ivory American Concert D. It's a sampled piano but the resonances are modeled. But to me, I hear no difference in this, versus the sampled resonances in the Ravenscorft 275. To me, these resonances are all the same. But the Ivory American Concert D is quite playable with a great tone. Dated, yes, but still good.
Posted By: Seb Clement Re: Modeling vs Sampling - 09/20/20 07:45 PM
Originally Posted by CyberGene
Seb, you’ve only tried sampled VST-s that are not among the good ones and don’t support repedaling or half-pedaling. You may want to try Garritan CFX with my own repedaling fix (there’s a thread about it). IMO it’s more playable than Pianoteq while also being drastically better sounding.
Noire and Ravenscroft claim to support half pedalling and repedaling - is this not the case? Does Garrison CFX support a continuous pedal with more than 3 values? If so, it's something I'd like to try!
Posted By: Gombessa Re: Modeling vs Sampling - 09/20/20 07:57 PM
Originally Posted by Seb Clement
I treat digital pianos as instruments in their own right - it's no good always holding them to the 'gold standard' of an acoustic grand piano - they're a separate instrument.

I think this perspective is more prevalent in those who own an acoustic kor who have ready access) as a primary instrument. In that case, the realism of the DP just isn't as high a priority because they always have the real thing. For those who have a DP as their only piano, it's oftentimes much more important that the DP be as accurate a simulation as possible (because they don't have any other choice).
Posted By: CyberGene Re: Modeling vs Sampling - 09/20/20 08:07 PM
Originally Posted by Seb Clement
Originally Posted by CyberGene
Seb, you’ve only tried sampled VST-s that are not among the good ones and don’t support repedaling or half-pedaling. You may want to try Garritan CFX with my own repedaling fix (there’s a thread about it). IMO it’s more playable than Pianoteq while also being drastically better sounding.
Noire and Ravenscroft claim to support half pedalling and repedaling - is this not the case? Does Garrison CFX support a continuous pedal with more than 3 values? If so, it's something I'd like to try!

I haven't personally played Noire or Ravenscroft but from the other user's opinions I've read, it seems half-pedaling and re-predaling are not very well scripted.

As to number of half-pedaling steps in Garritan CFX, I've just tested it and I can't notice any stepping, so it is either fully continuous or has enough steps to be unnoticeable. Maybe Jeff Hurchala (the developer, but he only rarely participates on this forums) can confirm this. The problem with Garritan is it can't be tested as Pianoteq and there's chance that you won't like it either and there are people who are still impressed more by Pianoteq. There's also some noise in Garritan CFX which may be more pronounced and annoying through particular headphones/speakers that seem to accentuate it... But if you're feeling adventurous, the lite versions is $70 AFAIK, which isn't something that will make a huge dent in your pocket. It lacks the distant microphones and thus makes it sounding dry, compared to the full version which captures the real ambiance of Abbey Road studios, but Lite is more or less representative of what you may expect in terms of playability and timbre quality and you can later upgrade to the Full version. The full version is really worth it IMO, even if for the added distant microphones on the classic mic perspective. I don't even use the other two perspectives that come with Full.
Posted By: David Lai Re: Modeling vs Sampling - 09/20/20 08:29 PM
The Ravenscroft can set a threshold where the half pedaling can take affect, and how this effect ends. So there are two midi values one can set. Personally, I think it's pretty good.
Posted By: JoeT Re: Modeling vs Sampling - 09/20/20 08:31 PM
Originally Posted by Del Vento
Originally Posted by JoeT
Please compare actual hardware instead of behind-the-curve PC implementations.

Which hardware?

The modeling team is represented by Roland etc.
The sampling team is represented by Yamaha, Kawai, Clavia Nord and Dexibell etc.

Their digital piano implementations are all way ahead of what is available as PC software, with the latter being stuck in the late 1990s to the early 2000s. I mean Yamaha solved half-pedaling in toy keyboards like NP-32, while expensive software still struggles with it like 1990s digitals did.
Posted By: JoeT Re: Modeling vs Sampling - 09/20/20 08:38 PM
Originally Posted by Gombessa
Originally Posted by Seb Clement
I treat digital pianos as instruments in their own right - it's no good always holding them to the 'gold standard' of an acoustic grand piano - they're a separate instrument.

I think this perspective is more prevalent in those who own an acoustic kor who have ready access) as a primary instrument. In that case, the realism of the DP just isn't as high a priority because they always have the real thing. For those who have a DP as their only piano, it's oftentimes much more important that the DP be as accurate a simulation as possible (because they don't have any other choice).

Digital instruments get measured them by what they advertise. There we have pictures of concert grands everywhere, regardless if brochures by hardware manufacturers or VSTi vendors. So that is the expectation they have to meet. I've never seen VSTi UI, which comes with the picture of a Casio keyboard. Even Pianoteq advertises "Steinway D".

So if they fall short, that is rightfully mentioned.
Posted By: Seb Clement Re: Modeling vs Sampling - 09/20/20 09:22 PM
Originally Posted by JoeT
Originally Posted by Del Vento
Originally Posted by JoeT
Please compare actual hardware instead of behind-the-curve PC implementations.

Which hardware?

The modeling team is represented by Roland etc.
The sampling team is represented by Yamaha, Kawai, Clavia Nord and Dexibell etc.

Their digital piano implementations are all way ahead of what is available as PC software, with the latter being stuck in the late 1990s to the early 2000s. I mean Yamaha solved half-pedaling in toy keyboards like NP-32, while expensive software still struggles with it like 1990s digitals did.
Are you claiming that the sampling technology is the current crop of digital pianos is better than software alternatives?

If so this is demonstrably absolute rubbish!
Posted By: magicpiano Re: Modeling vs Sampling - 09/20/20 10:11 PM
I think he refers just to the half-pedaling thing. Most digital pianos engines have good implementations of the half-pedaling.
Posted By: Del Vento Re: Modeling vs Sampling - 09/20/20 11:08 PM
Originally Posted by magicpiano
I think he refers just to the half-pedaling thing. Most digital pianos engines have good implementations of the half-pedaling.

But then most (all?) digital piano have such short sustains (probably to hide looping?) that it's kind of useless....
Posted By: Gombessa Re: Modeling vs Sampling - 09/20/20 11:20 PM
Originally Posted by JoeT
Digital instruments get measured them by what they advertise.

Only if they get measured by what they advertise. I think we've just seen here in this thread that not all people do so.

And that's fine. The point is that everyone has different criteria and standards, depending on their needs.
Posted By: navindra Re: Modeling vs Sampling - 09/21/20 05:55 AM
Originally Posted by Seb Clement
I also feel I can play with tone in the same sort of way that I can on an acoustic piano. You can bring in a subtle edge to the tone when required, and play with nuances between the left and right hands while playing that I simply can't recreate with most DPs. I do find that FF/FFF playing, especially in the bass of the piano DOES leave a lot to be desired in PianoTeq, but the tradeoff is still worth it for me.

I found a video from Phil Best on Youtube, that I can really relate to:

https://www.youtube.com/watch?v=j4SEgi5pADA

+1

As with any acoustic piano, playing and listening intently to the musical response is the key to unlocking the magic of Pianoteq. I feel like the musical depths of Pianoteq ought to be undeniable.

Objections to the core quality of the sound are certainly understandable, and even objectively quantifiable per the paper, but I've found even acoustics to have various flaws that I eventually overlook. I find the Pianoteq Bechstein, Steingraeber, U4, and others, to be quite enjoyable in practice.

Pianoteq instruments also vary in age -- generally the more recent or highly updated ones are superior.
Posted By: magicpiano Re: Modeling vs Sampling - 09/21/20 06:40 AM
Originally Posted by Del Vento
Originally Posted by magicpiano
I think he refers just to the half-pedaling thing. Most digital pianos engines have good implementations of the half-pedaling.

But then most (all?) digital piano have such short sustains (probably to hide looping?) that it's kind of useless....
That's not my experience with my Kawai DP (CN37). It has such a long and powerful sustain that you have to learn to use the pedal with great care if you want to get a nice sound. And if you want you can even make the sustain longer (or shorter) by changing a parameter in the options.
Posted By: TheodorN Re: Modeling vs Sampling - 09/21/20 10:21 AM
Is looped sustain as big a problem as many say it is? Can't we just drown it in reverb, LOL?
Posted By: FloRi89 Re: Modeling vs Sampling - 09/21/20 10:26 AM
Originally Posted by navindra
+1

As with any acoustic piano, playing and listening intently to the musical response is the key to unlocking the magic of Pianoteq. I feel like the musical depths of Pianoteq ought to be undeniable.

Objections to the core quality of the sound are certainly understandable, and even objectively quantifiable per the paper, but I've found even acoustics to have various flaws that I eventually overlook. I find the Pianoteq Bechstein, Steingraeber, U4, and others, to be quite enjoyable in practice.

Pianoteq instruments also vary in age -- generally the more recent or highly updated ones are superior.

The thing with Pianoteq is, if you play a comparison of let's say the Modern U and the U4, the Modern U sounds better. But that's kind of a sterile comparision that focuses on one very specific (albeit important) aspect. For me, for daily practice, Pianoteq is still the best solution in my current setup.
Posted By: U3piano Re: Modeling vs Sampling - 09/21/20 10:45 AM
Originally Posted by TheodorN
Is looped sustain as big a problem as many say it is? Can't we just drown it in reverb, LOL?

I'd think in theory it could be done very well, maybe even with some clever effect that makes the looped parts sound a bit different so the repeating of parts of the sound wouldn't be as noticable.

Still of course, unlooped is best. I think looped samples in 2020 is just not necessary, and a cheap solution.

The looped samples on my yamaha cp33 were terribly noticable.
Posted By: magicpiano Re: Modeling vs Sampling - 09/21/20 11:26 AM
Originally Posted by TheodorN
Is looped sustain as big a problem as many say it is? Can't we just drown it in reverb, LOL?
Is not a big problem, but of course it makes the piano sounds more digital, especially if the loop starts after less than 2 seconds from the attack. Reverb, chorus and resonance effects could make the loop less obvious, but you will lose the clarity and the character of the original sampled note.
If the looped part started after 10 seconds from the attack, we wouldn't even notice the presence of a loop. But that would require much more storage memory for the samples and we know most DP manufacturers are very "stingy" about the amount of storage memory for the samples. I think Dexibell is one of the few (maybe the only) DP manufacturers that makes sample-based piano engines with up to 15 seconds for the attack sound of the lower notes.
Posted By: JoeT Re: Modeling vs Sampling - 09/21/20 11:36 AM
Originally Posted by TheodorN
Is looped sustain as big a problem as many say it is? Can't we just drown it in reverb, LOL?

Unlooped samples is the only thing PC sample libraries had going for them, while they get everything else wrong. (We have unlooped hardware implementations, too. Not in the mainstream brands though.)

Originally Posted by "U3piano"
The looped samples on my yamaha cp33 were terribly noticable.

Well, TruePianos 1.0 and Pianoteq 1.0 sound terrible as well.

The CP33 was a 64-polyphony stage piano released in ancient history, when USB to HOST was bleeding edge. We have 2020 now and the CP88 is current.

Originally Posted by Seb Clement
Are you claiming that the sampling technology is the current crop of digital pianos is better than software alternatives?

For acoustic grand piano simulation: Indeed. And the hardware modeling alternatives are also better than what's available as software plugin. If you've had read the entire thread, you would have noticed the OP didn't care at all about comparing outdated software solutions for the umpth time, this is what he wrote coming out of the Kawai thread:

Originally Posted by Vikendios
There seems to be quite a few misunderstandings about the difference between Modeling and Sampling, which are quite different technologies for Digital Pianos. Today, full modeling is only offered by Roland and Viscount/Physis (originally developed by the latter for their organs[...]
Posted By: Del Vento Re: Modeling vs Sampling - 09/21/20 01:04 PM
Originally Posted by magicpiano
Originally Posted by Del Vento
Originally Posted by magicpiano
I think he refers just to the half-pedaling thing. Most digital pianos engines have good implementations of the half-pedaling.

But then most (all?) digital piano have such short sustains (probably to hide looping?) that it's kind of useless....
That's not my experience with my Kawai DP (CN37). It has such a long and powerful sustain that you have to learn to use the pedal with great care if you want to get a nice sound. And if you want you can even make the sustain longer (or shorter) by changing a parameter in the options.

Interesting. Admittedly I have not tried the latest crop of Kawai DP yet, but I was not expecting this from the Progressive Harmonic Imaging the CN37 and current CN39 have. Why? Because one of the pluses of the top of the line Harmonic Imaging XL (HI-XL) is the eXtra Long samples. I guess that's another reason to go to the Kawai dealer, in addition to testing the NV-5 laugh (even though I cannot afford it)

So let me rephrase my comment with

All digital pianos I have tried have such short sustains that the point is moot for slow playing, including the one I own, the NU1. And let me be even more specific, to avoid contributing to blanket statements which are plaguing this thread: stopwatch in hand, I measured sustain with PianoTeq (demo) being about the same as my golden era grand, and twice as long as the default NU1 engine.
Posted By: U3piano Re: Modeling vs Sampling - 09/21/20 01:13 PM
Originally Posted by "U3piano"
The looped samples on my yamaha cp33 were terribly noticable.

Originally Posted by "JoeT"
Well, TruePianos 1.0 and Pianoteq 1.0 sound terrible as well.

The CP33 was a 64-polyphony stage piano released in ancient history, when USB to HOST was bleeding edge. We have 2020 now and the CP88 is current.

Fair point, it is old tech.

Originally Posted by Seb Clement
Are you claiming that the sampling technology is the current crop of digital pianos is better than software alternatives?

Originally Posted by "JoeT"
For acoustic grand piano simulation: Indeed.

That would be an opinion, and not that I have tried all dp's, but I disagree. Especially after having experience with vsl synchron piano's, I highly doubt there are dp's that can compare, sampling technology wise.
Posted By: Del Vento Re: Modeling vs Sampling - 09/21/20 01:19 PM
Originally Posted by JoeT
Originally Posted by Seb Clement
Are you claiming that the sampling technology is the current crop of digital pianos is better than software alternatives?

For acoustic grand piano simulation: Indeed. And the hardware modeling alternatives are also better than what's available as software plugin. If you've had read the entire thread, you would have noticed the OP didn't care at all about comparing outdated software solutions for the umpth time

Well, threads evolve beyond what the OPs intend. This one had a quite lengthy discussion about BBQ, pizza and liquor... so I guess what we are discussing here is still on topic.

Anyway, interesting hearing about your point of view. I have not tried any of the software samplers myself (because they do not have demos, and I dislike shelling out non-refundable money for something I don't know if I like or not), so I can't express my judgment on that part. But I can tell that PianoTeq (demo) pianos sound marginally better than internal NU1 engine, when played via the same speakers (NU1 ones, with line-in input). To be more specific, by "marginally better" I mean "sounds closer to the sound of an acoustic piano, and less to a 'digital fake', particularly for pathological situation, such as parallel octaves, e.g. Hanon exercises (not that I care about that, just being clear on what I mean)". You may claim that the NU1 is not the NU1x, and that's fair....

On the other hand, if what you say is correct, why many people buy those software pianos when they own the "better" internal sound engine in their instruments which they use as a controller? Maybe just because of updates? E.g. should I have decided to buy PianoTeq, I could update to the current version for $39, whereas updating NU1 to NU1x is thousands? Or perhaps for the flexibility? With PianoTeq I could alter everything I want (e.g. voicing), whereas with the NU1 I'm stuck with the fairly bright Yamaha sound?

Last, but not least, can you provide some specific comparisons e.g. brand and model of piano being better (in what sense) than particular software piano? I'm especially interested if you can compare with Garritan.

Thanks!
Posted By: magicpiano Re: Modeling vs Sampling - 09/21/20 01:23 PM
Originally Posted by Del Vento
Interesting. Admittedly I have not tried the latest crop of Kawai DP yet, but I was not expecting this from the Progressive Harmonic Imaging the CN37 and current CN39 have. Why? Because one of the pluses of the top of the line Harmonic Imaging XL (HI-XL) is the eXtra Long samples. I guess that's another reason to go to the Kawai dealer, in addition to testing the NV-5 laugh (even though I cannot afford it)
It's true that the HI-XL has longer samples, but the duration of the sustain is independent of the sample length, because the piano engine can make the looped sustained part long/short almost as you want.
Posted By: matschulat Re: Modeling vs Sampling - 09/21/20 01:40 PM
Originally Posted by Del Vento
Matschulat, I totally agreed on the "this is better" part of the discussion, and I am sorry if any of my comments sounded like that (it was not my intention). In fact my experience is that every software or digital piano that I tried sounds horrible, even the worst (in tune) acoustic piano that I have ever tried sounded better, so I am trying to find something to like. Why? Because of the convenience of the digital instrument for certain things. I want to understand if I am doing something wrong, e.g. looking at something that is out of today's possibilities, while overlooking something that can be achieved today and I could be (at least partially) satisfied with. In this regard, I find this discussion extremely useful, e.g. my last exchange with Seb.

Yes, I must agree with you!

Originally Posted by Del Vento
As subjecting as it is, I think it would be useful if you can elaborate on this. I mean, you like Garritan for some specific reasons not like "my favorite color is green, period - there is no other reason". In what regard do you like Garritan better than PianoTeq and in what other regard the other way around? As you very clearly stated, in subjective terms. I think it's fine talking about subjective preference, like some preferring "bright" and other "mellow" in voicing an acoustic instrument, and then further diving deeper by saying "this technician have been able to use a special way to achieve very mellow, but not muffled at all sound". Hope I explained clearly what I am hoping to learn from you.

I think my subjective preference for Garritan instead of Pianoteq is because I find Garritan's piano tone more realist and beautiful. Also, its ambience (which is the real ambience of the studio where the piano was recorded) makes me feel like I'm in a hall playing, and it is also very suitable to recording. But, if I had the money, I'd buy Pianoteq also, since I realize the advantages it has, especially for practicing and getting a more accurate modeling of the physical behaviour of the piano. But I needed a VST for recording, so having the most recording-like sound was the priority for me.

It is subjective in the sense that different products are here to serve to different needs.
Posted By: JoeT Re: Modeling vs Sampling - 09/21/20 01:40 PM
Originally Posted by Del Vento
On the other hand, if what you say is correct, why many people buy those software pianos when they own the "better" internal sound engine in their instruments which they use as a controller?

The most controller in use by those guys is the Kawai VPC1 controller, which doesn't have any internal sound engine.

Quote
Last, but not least, can you provide some specific comparisons e.g. brand and model of piano being better (in what sense) than particular software piano? I'm especially interested if you can compare with Garritan.

Many people praising broken software like Garritan CFX are completely unfamiliar with Yamaha's own implementation of their own concert grand, which is far superior especially when it comes to "play-ability". Anything else would be pretty embarrassing for the market leader.

That usually comes up with the spiel that you have to sacrifice sound for play-ability or "treat digital pianos as their own specialy gifted instruments", while being completely oblivious to the fact, that technologies like Yamaha TransAcoustic or Kawai AURES exist.
Posted By: U3piano Re: Modeling vs Sampling - 09/21/20 01:52 PM
Originally Posted by JoeT
Many people praising broken software like Garritan CFX are completely unfamiliar with Yamaha's own implementation of their own concert grand, which is far superior especially when it comes to "play-ability".

Again, this would be an opinion, not a fact.

Also, it is interesting that some people with n1x's, nv10's and others, choose to use (broken?) software libraries such as garritans or vsl's, instead of using the build in piano sounds.
Posted By: Pete14 Re: Modeling vs Sampling - 09/21/20 02:04 PM
It’s also interesting that N1X owners tend to believe the N1X is better; whilst NV-10 owners are most certain that the NV-10 is better. I wonder why this happens?
Posted By: FloRi89 Re: Modeling vs Sampling - 09/21/20 03:33 PM
Originally Posted by Pete14
It’s also interesting that N1X owners tend to believe the N1X is better; whilst NV-10 owners are most certain that the NV-10 is better. I wonder why this happens?

And Garritan owners believe that's the best, while Pianoteq owners... It's almost if there is a bias for the product where you just left a substential amount of money. Someone should make a study about that...

Also MY totally unbiased opinion is that the Clavinova 675 with Pianoteq or the Ravenscroft 275 are UNDENIABLE the best things you can get. (And that's 100% not related to the fact that I own these products :D).
Posted By: MacMacMac Re: Modeling vs Sampling - 09/21/20 03:54 PM
I have the Garritan CFX. Do I think it's the best? No.
I have Pianoteq demo. Do I think it's the best? No.
I have the Grandeur. Do I think it's the best? No.
I have several dozen others. Do I think any of them are best? No ... with one exception. The Vintage D is my best.

So the notion that if someone owns X then X he deems X to be best ... doesn't hold for me.
Originally Posted by FloRi89
And Garritan owners believe that's the best, while Pianoteq owners... It's almost if there is a bias for the product where you just left a substential amount of money. Someone should make a study about that.
Posted By: Del Vento Re: Modeling vs Sampling - 09/21/20 04:00 PM
Originally Posted by FloRi89
Originally Posted by Pete14
It’s also interesting that N1X owners tend to believe the N1X is better; whilst NV-10 owners are most certain that the NV-10 is better. I wonder why this happens?

And Garritan owners believe that's the best, while Pianoteq owners... It's almost if there is a bias for the product where you just left a substential amount of money. Someone should make a study about that...

Also MY totally unbiased opinion is that the Clavinova 675 with Pianoteq or the Ravenscroft 275 are UNDENIABLE the best things you can get. (And that's 100% not related to the fact that I own these products :D).

This is partially obvious (one would not buy something they think is bad) and partially known psychological fact (coping with buyer's remorse, check it on wikipedia). The kind of answers I was looking with my question was the one matschulat wrote about. Thanks a lot for that matschulat!

Let me answer myself for others who might be interested. I bought my NU1 because I found it second hand from a friend at a steal price and it has much better action than the very old Kawai CA it replaced. I did not pay attention to the sound and I actually don't like too much the sound of either instruments (but I do liked that the sound from CA was somewhat customizable, whereas the NU1 sound is not). After the comments by Seb, I am actually not paying attention to the "I don't like the sound" anymore. and instead I am paying attention to the "what I can do better with this sound". And I'm already pleased with the (slight) improvements I'm seeing in my playing, because after all, that is what matters to me.

Thanks
Posted By: U3piano Re: Modeling vs Sampling - 09/21/20 04:02 PM
Well, I drive a Kia Picanto, but still I think a Porche Carrera would be better.
Posted By: Del Vento Re: Modeling vs Sampling - 09/21/20 04:02 PM
Originally Posted by MacMacMac
I have the Garritan CFX. Do I think it's the best? No.
I have Pianoteq demo. Do I think it's the best? No.
I have the Grandeur. Do I think it's the best? No.
I have several dozen others. Do I think any of them are best? No ... with one exception. The Vintage D is my best.

So the notion that if someone owns X then X he deems X to be best ... doesn't hold for me.

To make this answer useful to others (please see my previous posts for details), can you kindly elaborate on what do you like (or not) of each one (if anything) and why that one is best for you?
Posted By: Chordo24 Re: Modeling vs Sampling - 09/21/20 04:17 PM
Originally Posted by U3piano
Well, I drive a Kia Picanto, but still I think a Porche Carrera would be better.

So thinks Belastingdienst as well!
Posted By: FloRi89 Re: Modeling vs Sampling - 09/21/20 04:39 PM
Originally Posted by U3piano
Well, I drive a Kia Picanto, but still I think a Porche Carrera would be better.

Well most likely you are still convinced that this Kia was the best option for you and have reasons why that is so.

It’s called „post purchase rationalization“ and pretty well researched. Unless you have a case of buyers remorse, that also happens more often then one might think.

Funny enough though, a coping mechanism for buyers remorse is again rationalization of the purchase, so it can lead to the same thing: Overlooking of obvious flaws of a single purchase.
Posted By: MacMacMac Re: Modeling vs Sampling - 09/21/20 06:47 PM
With piano VSTs ... I generally don't suffer from post-purchase rationalization.

Nor do I have buyer's remorse ... at least not initially.

But I almost always have reservations about a VST ... which evolves into remorse ... which leads me to the next VST purchase. It's a GAS GAS GAS.
Posted By: EnoughBS Re: Modeling vs Sampling - 09/21/20 07:04 PM
Originally Posted by JoeT
Originally Posted by Del Vento
On the other hand, if what you say is correct, why many people buy those software pianos when they own the "better" internal sound engine in their instruments which they use as a controller?

The most controller in use by those guys is the Kawai VPC1 controller, which doesn't have any internal sound engine.

Quote
Last, but not least, can you provide some specific comparisons e.g. brand and model of piano being better (in what sense) than particular software piano? I'm especially interested if you can compare with Garritan.

Many people praising broken software like Garritan CFX are completely unfamiliar with Yamaha's own implementation of their own concert grand, which is far superior especially when it comes to "play-ability". Anything else would be pretty embarrassing for the market leader.

That usually comes up with the spiel that you have to sacrifice sound for play-ability or "treat digital pianos as their own specialy gifted instruments", while being completely oblivious to the fact, that technologies like Yamaha TransAcoustic or Kawai AURES exist.

You know JoeT, you seem to be talking a lot of **** and saying programs like Garritan are "broken", but I don't see you giving any details to validate that contrarian stance.
Posted By: FloRi89 Re: Modeling vs Sampling - 09/21/20 08:04 PM
Originally Posted by MacMacMac
With piano VSTs ... I generally don't suffer from post-purchase rationalization.

Nor do I have buyer's remorse ... at least not initially.

But I almost always have reservations about a VST ... which evolves into remorse ... which leads me to the next VST purchase. It's a GAS GAS GAS.

That will explain the amount of VSTs you listed before ^^. Really not one that could keep your fancy?
Posted By: EB5AGV Re: Modeling vs Sampling - 09/21/20 08:07 PM
Originally Posted by MacMacMac
With piano VSTs ... I generally don't suffer from post-purchase rationalization.

Nor do I have buyer's remorse ... at least not initially.

But I almost always have reservations about a VST ... which evolves into remorse ... which leads me to the next VST purchase. It's a GAS GAS GAS.

Mmmmm... That sounds familiar to me. I have bought quite a lot of piano VSTs lately (mostly due to positive comments on this forum wink ). My excuse is that they don't take space and I get them always on discounted sales or educational prices... For me it is a valid excuse and it is lovely to be able to choose among a lot of different sounds thumb
Posted By: MacMacMac Re: Modeling vs Sampling - 09/21/20 08:38 PM
I spent lots of time with the Grandeur back when. I had set aside the Vintage D because I did not have adequate control.

But I eventually went back to the Vintage D (because it sounds better), and that forced me to learn better control.
Originally Posted by FloRi89
Really not one that could keep your fancy?

Years earlier I really liked a couple of Vienna (Bosie) pianos, and the Kawai EX Pro. But the Bosies are a bit thin, and that Kawai lacked air.

As for the rest ... I'm starting to get back with the Garritan CFX. I'm having pedaling problems that I've not been able to solve.

I have a Fazioli, too ... but it's much too ordinary. It lacks character.

The old Berlin Grand is still installed, but the other Steinways have it beat.

I have the Giant, the Maverick, and the Gentleman, but I don't use them much anymore.

I long ago uninstalled the Bechstein, 8DIO Legacy, Old Black Grand, Sampletekk White Grand, Alicia's Keys, and a few others.
Those ranged for so-so to major stink, especially Alicia's Keys.

And finally ... the Pianoteq demo is still installed. It serves as a no-cost alternative to Ipecac.
Posted By: FloRi89 Re: Modeling vs Sampling - 09/22/20 08:17 AM
What about the Modern U? I‘m surprised that you don’t own that. Lot’s of people seem to really like that one.
Posted By: MacMacMac Re: Modeling vs Sampling - 09/22/20 09:12 AM
The G.A.S. has settled down a bit over the past year or so. The CFX was my only recent acquisition. And I don't think I'll be buying more.
That's because I have several good and excellent packages now, and I don't need any more of them.
Posted By: EB5AGV Re: Modeling vs Sampling - 09/22/20 12:21 PM
Originally Posted by MacMacMac
The G.A.S. has settled down a bit over the past year or so. The CFX was my only recent acquisition. And I don't think I'll be buying more.
That's because I have several good and excellent packages now, and I don't need any more of them.

You should try the Modern U, is gorgeous!

I took the plunge on it plus Ravenscroft 275 and also on Signature Grand, Garritan CFX and some extra Pianoteq instruments this Summer, all at discounted prices (my excuse eek) . I am done for some time!
Posted By: QuasiUnaFantasia Re: Modeling vs Sampling - 09/22/20 01:01 PM
Originally Posted by FloRi89
And Garritan owners believe that's the best, while Pianoteq owners... It's almost if there is a bias for the product where you just left a substential amount of money. Someone should make a study about that...

Hmmm ... I have both Garritan CFX and Pianoteq, and the latter is much more practical in use (e.g. it is fabulous for recording MIDI, and it loads really rapidly), but the former sounds much better. Neither is perfect, and I'm still looking for that perfect VI (and not expecting to find it any time soon).
Posted By: mcontraveos Re: Modeling vs Sampling - 09/22/20 09:32 PM
Originally Posted by MacMacMac
I have a Fazioli, too ... but it's much too ordinary. It lacks character.
Real Faziolis share this limitation as well.

Quote
And finally ... the Pianoteq demo is still installed. It serves as a no-cost alternative to Ipecac.
This is insulting; my family is a pillar in the ipecac syrup industry and we resent our product being compared to Pianoteq.
Posted By: meghdad Re: Modeling vs Sampling - 09/23/20 07:58 AM
Can someone knowledgeable explain to me briefly what are the sonic advantages and disadvantages of the PCM sampling method compared to other sampling methods and modeling?
Posted By: CyberGene Re: Modeling vs Sampling - 09/23/20 08:08 AM
Originally Posted by meghdad
Can someone knowledgeable explain to me briefly what are the sonic advantages and disadvantages of the PCM sampling method compared to other sampling methods and modeling?

PCM is just the technical term for WAV files. It means the sound is sampled periodically (commonly 44100 times a second). I guess there are no other types of sampling. A possible difference could be the usage of lossless (FLAC) or lossy (MP3/AAC, etc.) compression to store the samples.
Posted By: meghdad Re: Modeling vs Sampling - 09/23/20 08:17 AM
Originally Posted by CyberGene
I guess there are no other types of sampling. A possible difference could be the usage of lossless (FLAC) or lossy (MP3/AAC, etc.) compression to store the samples.
So if there's no other type of sampling, does that mean that for example the difference between Pure CF Yamaha sound engine and simply termed PCM Sampling of Korg only boils down to the manual process of capturing the sound (and perhaps further post-processing)?

I mean, essentially there shouldn't be much of a difference between them then, if at all right? It's not rocket science and it's old technology right?
Posted By: CyberGene Re: Modeling vs Sampling - 09/23/20 08:25 AM
Originally Posted by meghdad
Originally Posted by CyberGene
I guess there are no other types of sampling. A possible difference could be the usage of lossless (FLAC) or lossy (MP3/AAC, etc.) compression to store the samples.
So if there's no other type of sampling, does that mean that for example the difference between Pure CF Yamaha sound engine and simply termed PCM Sampling of Korg only boils down to the manual process of capturing the sound (and perhaps further post-processing)?

I mean, essentially there shouldn't be much of a difference between them then, if at all right? It's not rocket science and it's old technology right?

All sampling methods involve recording the original piano as WAV files, multiple velocities, all keys, pedal up and down, release samples, etc.

When they mention some fancy names such as "pure CF sampling", "supernatural", etc. they (apart from the obvious marketing BS) mean that there may be some clever technology in the reproduction of the samples, such as to make looping less obvious, sample interpolation between velocity layers to make switch less obvious, round-robin of samples for more variety, etc. However I am yet to see a company reveal any of these beside vague hints, so it's ultimately up to how well the piano plays and sounds to you.
Posted By: JoeT Re: Modeling vs Sampling - 09/23/20 11:02 AM
Originally Posted by CyberGene
Originally Posted by meghdad
Can someone knowledgeable explain to me briefly what are the sonic advantages and disadvantages of the PCM sampling method compared to other sampling methods and modeling?

PCM is just the technical term for WAV files. It means the sound is sampled periodically (commonly 44100 times a second). I guess there are no other types of sampling.

You guessed wrong. PCM is not a technical term for RIFF WAVE files - the official name for the file format. This is due to the fact, that WAV files are containers, which can store all kinds of multi-channel audio formats including compressed audio like MP3 and AAC (which employ their own MPEG container formats, but you don't have to use those).

PCM is a term for Pulse-Code Modulation, which is one of several methods of sampling, which got well known with the introduction of the CD Digital Audio. But it's not the only one, for example the SACD uses Delta-sigma Modulation, which is Pulse-Density Modulation (PDM).

So when Korg states they used PCM sampling, then they just stated they used the most wide-spread sampling method, because their hardware processes PCM data.
Posted By: MacMacMac Re: Modeling vs Sampling - 09/23/20 11:19 AM
I can appreciate your curiosity ...
Originally Posted by meghdad
Can someone knowledgeable explain to me briefly what are the sonic advantages and disadvantages of the PCM sampling method compared to other sampling methods and modeling?
But consider this: The underlying tech doesn't really matter. Here's why:

1. The piano takes from your fingers and gives to your ears. In between those two points the piano does some work. Does it matter how? Or can we simply judge (a) how the piano feels and (b) how the piano sounds? And maybe (c) how the piano looks? To me these are the only things that matter.

2. Descriptions of "what's under the hood" are often distorted by misunderstanding, and especially by marketing jargon, nonsense, rubbish, balderdash, gibberish, blarney, guff, blather, claptrap, and gobbledygook. Stuff to be ignored. smile
Posted By: CyberGene Re: Modeling vs Sampling - 09/23/20 11:36 AM
Originally Posted by JoeT
Originally Posted by CyberGene
Originally Posted by meghdad
Can someone knowledgeable explain to me briefly what are the sonic advantages and disadvantages of the PCM sampling method compared to other sampling methods and modeling?

PCM is just the technical term for WAV files. It means the sound is sampled periodically (commonly 44100 times a second). I guess there are no other types of sampling.

You guessed wrong. PCM is not a technical term for RIFF WAVE files - the official name for the file format. This is due to the fact, that WAV files are containers, which can store all kinds of multi-channel audio formats including compressed audio like MP3 and AAC (which employ their own MPEG container formats, but you don't have to use those).

PCM is a term for Pulse-Code Modulation, which is one of several methods of sampling, which got well known with the introduction of the CD Digital Audio. But it's not the only one, for example the SACD uses Delta-sigma Modulation, which is Pulse-Density Modulation (PDM).

So when Korg states they used PCM sampling, then they just stated they used the most wide-spread sampling method, because their hardware processes PCM data.

I explained to the guy who asked about PCM that PCM = WAVE file because that's how I believe he can understand it and that's what most people understand when they use WAV files smile

Regarding the sigma-delta 1-bit modulation, I haven't heard a digital piano manufacturer that use it, which is why I *guess* there are no other sampling methods *currently in use*. The sigma-delta sampling is pretty difficult (if not impossible) for further DSP processing such as EQ-ing and even for sample manipulation such as looping and gain-control over the looped section and the other piano-related audio manipulations related to sample-based piano playback which is why I believe there are no usages of sigma-delta. If you know of a DP-manufacturer or VST sample-based piano that uses sigma-delta, I'd be interested to learn.

As an offtopic, as always, when you try to explain something in a simpler way to someone who apparently isn't familiar with audio concepts, there will be the smart-pants guy who will correct you... wink Hopefully meghad now understands what a container is and what a codec is and all that stuff laugh
Posted By: meghdad Re: Modeling vs Sampling - 09/23/20 12:28 PM
Well thanks! But guess what, I am a software developer by day and piano learner by night! laugh So I am familiar with the containers and formats and what not. I just needed to know if there are other piano sampling methods which is why I asked a comparative question. smile
Posted By: Pete14 Re: Modeling vs Sampling - 09/23/20 03:56 PM
Have you considered developing software by day that could aid in learning the piano by night?
Posted By: Frédéric L Re: Modeling vs Sampling - 09/23/20 05:09 PM
Originally Posted by meghdad
Well thanks! But guess what, I am a software developer by day and piano learner by night! laugh So I am familiar with the containers and formats and what not. I just needed to know if there are other piano sampling methods which is why I asked a comparative question. smile
You have an answer about AWM (Advanced Wave Memory of Yamaha) :

http://www.motifator.com/index.php/forum/viewthread/454955/

But as a proprietary technology, you won’t know much about the internals.
Posted By: MacMacMac Re: Modeling vs Sampling - 09/23/20 07:29 PM
@FL: That link from 2011 sounds a bit like this one in 2020.

Q: What is AWM? What is AWM2?
A: They are marketing mumbledy-bum.

Q: What's the technology?
A1: Why does it matter?
A2: But I really want to know!
A3: Speculation has it that bits and bytes and blah-blah-blah mumbledy-bum. (All guesswork. No facts.)

My conclusion: Just listen.

If it sounds good, does it matter how many bits it has?
If it sounds bad, does it matter how many bits it has?
Posted By: JoeT Re: Modeling vs Sampling - 09/23/20 08:29 PM
MacMacMac You know you are just posting noise into threads?
Posted By: magicpiano Re: Modeling vs Sampling - 09/23/20 09:22 PM
Originally Posted by MacMacMac
[...]If it sounds bad, does it matter how many bits it has?
Yes, so that you can insult the manufacturer a "bit" more! laugh
Posted By: Frédéric L Re: Modeling vs Sampling - 09/24/20 05:24 PM
@MacMacMac : I have just answered the question « I just needed to know if there are other piano sampling methods which is why I asked a comparative question. ».

You reply to an other one : « does it matter ? ».

And you are right : only the result (what we hear) counts.
Posted By: U3piano Re: Modeling vs Sampling - 09/24/20 07:49 PM
Well, sure the results are most important, but say you are a potential buyer for a car and after a test drive you find it drives well, you still want to look under the hood, right?

With piano vst's you usually can't even take that test drive, you can only watch when someone else takes a drive, so that look under the hood becomes a bit more important. ☺️
Posted By: meghdad Re: Modeling vs Sampling - 09/24/20 08:34 PM
I believe it could be important particularly to a newbie who might not possess strong musical ears to hear the subtleties that become important to him in advanced stages of learning.
Posted By: Frédéric L Re: Modeling vs Sampling - 09/25/20 05:49 AM
The spec figures doesn’t tell all thing. Let’s take the number of velocity layers. In a first approximation, the more velocity layers there are, the harder it will be to detect the gap between them. But some virtual pianos are so badly sampled that you can spot some gaps even with 10-18 layers, when some are OK with only 8 layers.
Posted By: U3piano Re: Modeling vs Sampling - 09/25/20 06:00 AM
Originally Posted by Frédéric L
The spec figures doesn’t tell all thing. Let’s take the number of velocity layers. In a first approximation, the more velocity layers there are, the harder it will be to detect the gap between them. But some virtual pianos are so badly sampled that you can spot some gaps even with 10-18 layers, when some are OK with only 8 layers.

Very true.
Posted By: JoeT Re: Modeling vs Sampling - 09/25/20 07:14 AM
Originally Posted by Frédéric L
The spec figures doesn’t tell all thing. Let’s take the number of velocity layers. In a first approximation, the more velocity layers there are, the harder it will be to detect the gap between them. But some virtual pianos are so badly sampled that you can spot some gaps even with 10-18 layers, when some are OK with only 8 layers.

Layer switching still being a problem in PC software tells us how far behind the curve it is. I guess Kontakt isn't very good at "Harmonic Imaging XXXXXL". wink
Posted By: U3piano Re: Modeling vs Sampling - 09/25/20 07:32 AM
Originally Posted by JoeT
Layer switching still being a problem in PC software tells us how far behind the curve it is. I guess Kontakt isn't very good at "Harmonic Imaging XXXXXL". wink

I think it has nothing to do with PC software in general, but with specific qualities of different vst's and how well they are created.
Posted By: Burkey Re: Modeling vs Sampling - 09/25/20 07:35 AM
Originally Posted by Frédéric L
The spec figures doesn’t tell all thing. Let’s take the number of velocity layers. In a first approximation, the more velocity layers there are, the harder it will be to detect the gap between them. But some virtual pianos are so badly sampled that you can spot some gaps even with 10-18 layers, when some are OK with only 8 layers.
According to Mozart there are only 8 velocities:
ppp, pp, p, mp, mf, f, ff, fff

If it's good enough for Mozart, and good enough for me, then dammit it's good enough for you too!
Posted By: JoeT Re: Modeling vs Sampling - 09/25/20 08:21 AM
Originally Posted by U3piano
Originally Posted by JoeT
Layer switching still being a problem in PC software tells us how far behind the curve it is. I guess Kontakt isn't very good at "Harmonic Imaging XXXXXL". wink

I think it has nothing to do with PC software in general, but with specific qualities of different vst's and how well they are created.

It has very much to do with the PC software used for sample playback, which is Kontakt in the most prominent cases, with a few non-sample-based exceptions like Pianoteq.

Kontakt is the reason why VSTi regularly fail to get both layer switching, half-pedaling and re-pedaling right. Something sample-based digital pianos have solved two decades ago, even when they come with only three sample layers.
Posted By: U3piano Re: Modeling vs Sampling - 09/25/20 08:45 AM
Originally Posted by JoeT
It has very much to do with the PC software used for sample playback, which is Kontakt in the most prominent cases, with a few non-sample-based exceptions like Pianoteq.

Kontakt in the most prominent cases? I don't know about other instruments, but for piano vst's I don't think so.

There is a very nice list of some of the best (if not the best) piano libary creators that do not use kontakt: For example: Garritan, VSL, Synthogy, VI labs... These all have very nice piano's, without layer switching problems, so let's not make a PC software problem out of it. If it's a Kontakt problem, it's a Kontakt problem.
Posted By: MacMacMac Re: Modeling vs Sampling - 09/25/20 09:58 AM
I use Kontakt instruments ... and I don't hear any layering problems at all.

But I did hear it quite plainly with the native piano sounds, especially in the C4 and C5 octaves.
This is a three-layer Clav from 2005.
It seems that's something that digital pianos did NOT solve two decades ago, even when they come with only three sample layers.
Posted By: Del Vento Re: Modeling vs Sampling - 09/25/20 11:31 AM
Originally Posted by Burkie
Originally Posted by Frédéric L
The spec figures doesn’t tell all thing. Let’s take the number of velocity layers. In a first approximation, the more velocity layers there are, the harder it will be to detect the gap between them. But some virtual pianos are so badly sampled that you can spot some gaps even with 10-18 layers, when some are OK with only 8 layers.
According to Mozart there are only 8 velocities:
ppp, pp, p, mp, mf, f, ff, fff

If it's good enough for Mozart, and good enough for me, then dammit it's good enough for you too!

It might be good enough for digital pianos (software or hardware implemented) if interpolation is well implemented -- for that I agree with you.

Yet from a musical perspective that is plain wrong. Eight dynamics is what a musician would notate. But if you play with only 8 dynamic levels you are going to sound horrible. Often (but not always) you have to be much more gradual in your crescendo and diminuendo between levels. And always, in every single musical phrase you have to shape it with microdynamics even smaller than crescendos and diminuendos. Otherwise you are gonna sound like a typewriter, Doing this is already a difficult process in its own to learn, if the instrument does not respond correctly it's pretty much impossible and you will never learn one of the most beautiful aspects of the music.

Perhaps this is the point JoeT is trying to make (with his naysayer attitude which most of us do not like). and that is definitely the point that some people (e.g. Phil Best) make about PianoTeq playability
Posted By: U3piano Re: Modeling vs Sampling - 09/25/20 11:37 AM
Meh, Mozart is outdated!



*hides*
Posted By: CyberGene Re: Modeling vs Sampling - 09/25/20 12:38 PM
But Pianoteq is not unequivocally regarded as more playable. I don't know where that comes from, I often see it mentioned on the forum, but I am convinced it's just a user opinion or a matter of taste. I've read many times in the past a common misconception about how sampled pianos have only three dynamic layers, hence Pianoteq must be more playable because it has 127. That's not true, because even the worst sampled pianos have 127 velocity layers too. And so on, and so on. A misunderstanding like that can easily become a "fact" how Pianoteq is more playable.

We don't have quantitive definition of "playability".

However if judging by how well a piano responds to keyboard control and calling that "playability", then IMO Garritan CFX responds better than Pianoteq, hence for me it's more playable. I feel I have finer gradation of the timbre and the dynamics and they are more linear and closer to a real piano than Pianoteq. While I believe one can tweak Pianoteq in any possible way (MIDI velocity mapping and note by note editing, etc.), I haven't managed to ever make it so smooth and so I usually just give up.
Posted By: Del Vento Re: Modeling vs Sampling - 09/25/20 01:05 PM
Originally Posted by CyberGene
But Pianoteq is not unequivocally regarded as more playable. I don't know where that comes from, I often see it mentioned on the forum, but I am convinced it's just a user opinion or a matter of taste. I've read many times in the past a common misconception about how sampled pianos have only three dynamic layers, hence Pianoteq must be more playable because it has 127. That's not true, because even the worst sampled pianos have 127 velocity layers too. And so on, and so on. A misunderstanding like that can easily become a "fact" how Pianoteq is more playable.

We don't have quantitive definition of "playability".

However if judging by how well a piano responds to keyboard control and calling that "playability", then IMO Garritan CFX responds better than Pianoteq, hence for me it's more playable. I feel I have finer gradation of the timbre and the dynamics and they are more linear and closer to a real piano than Pianoteq. While I believe one can tweak Pianoteq in any possible way (MIDI velocity mapping and note by note editing, etc.), I haven't managed to ever make it so smooth and so I usually just give up.

Thanks for sharing this CG. I was just reporting what I have heard from others (e.g. Phil Best) without much experience on my own. In fact on my own I was only looking at how realistic the piano sounded, and be extremely disappointed with everything I tried. Only a week ago I started paying attention to playability, and only with PianoTeq (demo) and internal NU1 sound. Regarding Garritan CFX, do you mean both the Lite and the Full version or just the latter?
Thanks again
Posted By: CyberGene Re: Modeling vs Sampling - 09/25/20 01:10 PM
I like both versions of Garritan CFX but the added distant mics of the full version really make it stand out.
Posted By: Frédéric L Re: Modeling vs Sampling - 09/25/20 03:02 PM
Originally Posted by Burkie
According to Mozart there are only 8 velocities:
ppp, pp, p, mp, mf, f, ff, fff

If it's good enough for Mozart, and good enough for me, then dammit it's good enough for you too!

With velocity layers gaps, if we play near a gap, because of the dispersion, we have half the notes, mp, the other half mf... And we hear something disconnected with what we are trying to play.
Posted By: FloRi89 Re: Modeling vs Sampling - 09/25/20 03:05 PM
Originally Posted by CyberGene
But Pianoteq is not unequivocally regarded as more playable. I don't know where that comes from, I often see it mentioned on the forum, but I am convinced it's just a user opinion or a matter of taste. I've read many times in the past a common misconception about how sampled pianos have only three dynamic layers, hence Pianoteq must be more playable because it has 127. That's not true, because even the worst sampled pianos have 127 velocity layers too. And so on, and so on. A misunderstanding like that can easily become a "fact" how Pianoteq is more playable.

We don't have quantitive definition of "playability".

However if judging by how well a piano responds to keyboard control and calling that "playability", then IMO Garritan CFX responds better than Pianoteq, hence for me it's more playable. I feel I have finer gradation of the timbre and the dynamics and they are more linear and closer to a real piano than Pianoteq. While I believe one can tweak Pianoteq in any possible way (MIDI velocity mapping and note by note editing, etc.), I haven't managed to ever make it so smooth and so I usually just give up.

I'm not such a good player, but I can easily tell that the playability of Pianoteq is a lot closer to our real grands then the internal sounds of my Clavinova 675. Probably the most notable difference is when you switch the Yamaha grand to "silent" because then you have the same action, but a sampled piano. My dad is annoying the neighbours though because he claims that in silent mode the piano won't respond to his "emotions" (he can't describe it better then that) as much as a real piano. Something he claims that is much more present for Pianoteq (even though still to far off for his taste, so he stays with the acoustic for probably the rest of his life).

As for myself, I find it much easier to play with expression on the acoustic and Pianoteq as compared to the Clavinova sampled CFX. I can't speak for any other VSTs, it might be different but it's noticable in the VSTs I have access to (YAMAHA default CFX, Pianoteq, Ravenscroft 275). My teacher also recognized this and encourages me to use Pianoteq because my playing sounds a lot better with it.

Now this is a lot of personal opinions here, and I can't create any scientific reasoning for that, but to me, my family and my teacher there is a definite difference in playability.
Posted By: CyberGene Re: Modeling vs Sampling - 09/25/20 03:58 PM
You can find both amateurs and advanced pianists liking either sampled pianos or Pianoteq. It’s all about personal taste. The most dividing point seems to be an artificial (synthetic, metallic) character in Pianoteq which Pianoteq proponents can’t hear or are simply not bothered by. Putting that aside, Pianoteq is better in any respect compared to sample-based pianos. However some people (including me) just can’t ignore the synthetic character that spoils the pleasure of playing a piano for them and it’s a deal breaker.
Posted By: Gombessa Re: Modeling vs Sampling - 09/25/20 04:14 PM
Originally Posted by CyberGene
You can find both amateurs and advanced pianists liking either sampled pianos or Pianoteq. It’s all about personal taste. The most dividing point seems to be an artificial (synthetic, metallic) character in Pianoteq which Pianoteq proponents can’t hear or are simply not bothered by. Putting that aside, Pianoteq is better in any respect compared to sample-based pianos. However some people (including me) just can’t ignore the synthetic character that spoils the pleasure of playing a piano for them and it’s a deal breaker.

I pretty much agree entirely here, except for the bolded part, I'm not sure if that's what you intended? If Pianoteq didn't have that metallic/synthetic ring, I would definitely like it a lot more, but I don't think it wins on all other counts. Garritan CFX still has that wonderfully organic ambient/remote mic perspective that creates an actual, realistic reverb, this is IMO head and shoulders over any synthetic reverb, whether on a sampled or modeled VST or hardware DP engine.
Posted By: EPW Re: Modeling vs Sampling - 09/25/20 04:18 PM
I think it is the connection I have while playing with Pianoteq that makes the difference for me.
Now for CyberGene it is Garritan which is perfectly fine. We all have different tastes. Use what works for you and get on with practicing/playing the PIANO smile
Posted By: CyberGene Re: Modeling vs Sampling - 09/25/20 04:44 PM
Originally Posted by Gombessa
Originally Posted by CyberGene
You can find both amateurs and advanced pianists liking either sampled pianos or Pianoteq. It’s all about personal taste. The most dividing point seems to be an artificial (synthetic, metallic) character in Pianoteq which Pianoteq proponents can’t hear or are simply not bothered by. Putting that aside, Pianoteq is better in any respect compared to sample-based pianos. However some people (including me) just can’t ignore the synthetic character that spoils the pleasure of playing a piano for them and it’s a deal breaker.

I pretty much agree entirely here, except for the bolded part, I'm not sure if that's what you intended? If Pianoteq didn't have that metallic/synthetic ring, I would definitely like it a lot more, but I don't think it wins on all other counts. Garritan CFX still has that wonderfully organic ambient/remote mic perspective that creates an actual, realistic reverb, this is IMO head and shoulders over any synthetic reverb, whether on a sampled or modeled VST or hardware DP engine.

Actually I agree with you on that. I have entirely forgotten about reverb. I was thinking about things like pedaling, half-pedaling and resonances that are spot-on.

But yes, the real ambiance in Garritan CFX (and other libraries with distant mics) is another area which is lacking in modeled pianos (and any other digital/software piano with algorithmic reverb).
Posted By: Harpuia Re: Modeling vs Sampling - 09/25/20 06:56 PM
Regarding to the playability and how the VST responds to the touch, it is all tweak-able by the velocity curve. From my experience, different VSTs really need a different velocity curve if using the same keyboard. And if you switch to another keyboard, you need another velocity curve. I do think that this affects a lot more than whether it is sampled or modeled. For example, 20 velocity layers in Garritan CFX is already a lot but how to map these 20 layers is the key.

Now going back to the original question, for digital pianos and VSTs, we have a few different approaches:
1. DP manufactures with about 5 velocity layers with interpolations
2. Most sampled based VSTs with 8-36 layers that people usually like here, but without interpolations (Ravenscroft, Ivory II ACD, Garritan CFX, Embertone Walker 1955, etc.)
3. Brute force sampled VSTs with huge layers (~70 layers in VSL Synchron Pianos, and ~100 layers in VSL Vienna Imperial)
4. "fully modeled" solutions such as Pianoteq and Roland. Theoretically it should have 127 layers.

A lot of people noticed that these numbers are only on paper and more velocity layers really did not mean better playing experience. We also have a lot of variations here, such as how well those layers interpolate, and how well is the model in Pianoteq. How do we map those 70 layers of timbers in VSL Synchron Pianos to the 127 midi levels, how to map midi levels with loudness, and how to map "hammer velocity" to a midi level.

My personal impression is that interpolations tend to make the sound less detailed and lifeless, and fully model solution is not able to create a realistic attack sound at this point though the versatility of the sound is good. I do think that the "brute force" sampled solution by VSL has its edge at this point if sampled, implemented and scripted well.

In terms of pedaling, Pianoteq wins hands down followed by hardware DPs. Software VSTs are lagging behind. In terms of reverb, I agree with Gombassa and CyberGene that the real reverb is so much better than the artificial one, and it actually hides some "defects" of the sampled VSTs (sympathetic resonance, pedalling, note release, etc)
Posted By: navindra Re: Modeling vs Sampling - 09/25/20 06:58 PM
Originally Posted by CyberGene
Originally Posted by Gombessa
Garritan CFX still has that wonderfully organic ambient/remote mic perspective that creates an actual, realistic reverb, this is IMO head and shoulders over any synthetic reverb, whether on a sampled or modeled VST or hardware DP engine.

Actually I agree with you on that. I have entirely forgotten about reverb. I was thinking about things like pedaling, half-pedaling and resonances that are spot-on.

But yes, the real ambiance in Garritan CFX (and other libraries with distant mics) is another area which is lacking in modeled pianos (and any other digital/software piano with algorithmic reverb).


Reverb is definitely a challenge, however Pianoteq doesn't do so bad, from a piano player perspective.

If you want Pianoteq to sound really bad -- turn off reverb. That also helps demonstrate how well Modartt is actually doing. You don't have to care about anything other than the default reverb settings while playing.

For production purposes, you can reposition mics easily and use an external reverb, but that is optional, depending on your mix and depending on what you want to achieve in your recording.

I think Pianoteq mic positioning + Valhalla Room does a remarkable job here, for example:



I definitely hope that reverb is something that's revisited in Pianoteq 7 though, as it could have a tremendous payoff in how Pianoteq is perceived by critics.
Posted By: Gombessa Re: Modeling vs Sampling - 09/25/20 09:35 PM
I don't think the reverb is "bad" or unpleasant sounding on most any modern VST/hardware implementation. It's just that to my ears, nothing holds a candle to Garritan CFX (and presumably some other far-mic'd VSTs like VSL) in terms of actually sounding authentic/realistic. CFX really sounds like you're in and being recorded in the space, as opposed to just having a nice reverb effect applied to the tone.

The downside of course is that you can't really modify/customize that authenticity if you're not a fan of it or want a different ambiance. IMO, it does this one thing really, really well, but of course ya gotta like that one thing to have it be worth it.
Posted By: U3piano Re: Modeling vs Sampling - 09/25/20 09:56 PM
Originally Posted by Gombessa
(and presumably some other far-mic'd VSTs like VSL)

That's the impression i always got of vsl vst's as well, because in most of their demo's you hear alot of far mics, I don't know why they do that. To me that echoing sound isn't nice at all.


I was happy to find the close mics are close, as they should be. And then you have mid mics, which are... well, mid, as they should be. A vsl piano doesn't have to sound anything like these far mic'ed deno's.
Posted By: JoeT Re: Modeling vs Sampling - 09/25/20 10:33 PM
Originally Posted by Gombessa
Originally Posted by CyberGene
You can find both amateurs and advanced pianists liking either sampled pianos or Pianoteq. It’s all about personal taste. The most dividing point seems to be an artificial (synthetic, metallic) character in Pianoteq which Pianoteq proponents can’t hear or are simply not bothered by. Putting that aside, Pianoteq is better in any respect compared to sample-based pianos. However some people (including me) just can’t ignore the synthetic character that spoils the pleasure of playing a piano for them and it’s a deal breaker.

I pretty much agree entirely here, except for the bolded part, I'm not sure if that's what you intended? If Pianoteq didn't have that metallic/synthetic ring, I would definitely like it a lot more, but I don't think it wins on all other counts. Garritan CFX still has that wonderfully organic ambient/remote mic perspective that creates an actual, realistic reverb, this is IMO head and shoulders over any synthetic reverb, whether on a sampled or modeled VST or hardware DP engine.

Pianoteq is a great modeled harpsichord. But Pianoforte modeling, well...
Posted By: navindra Re: Modeling vs Sampling - 09/25/20 10:52 PM
Originally Posted by JoeT
Pianoteq is a great modeled harpsichord. But Pianoforte modeling, well...

Which harpsichord do you recommend in Pianoteq? There are several options but I haven't really explored them.

Pianoteq even lets you play them with pianoforte-like dynamic expression -- might make for something interesting if you find the right music to play.
Posted By: Frédéric L Re: Modeling vs Sampling - 09/25/20 10:56 PM
Originally Posted by Harpuia
2. Most sampled based VSTs with 8-36 layers that people usually like here, but without interpolations (Ravenscroft, Ivory II ACD, Garritan CFX, Embertone Walker 1955, etc.)
According to the Ivory page :

“Sample Interpolation Technology used for ultrasmooth velocity and note transitions”

Then Synthogy Ivory’s pianos are among the few VST (the only?) which use interpolation and avoid any velocity layers gap.

Kontakt enables velocity crossfade (see https://www.adsrsounds.com/kontakt-tutorials/creating-key-and-velocity-crossfades-old/ ) but this doesn’t mean many Kontakt librairies use this feature. I suppose we can even have phase cancellation if samples are not well aligned.
Posted By: JoeT Re: Modeling vs Sampling - 09/25/20 11:05 PM
Originally Posted by navindra
Originally Posted by JoeT
Pianoteq is a great modeled harpsichord. But Pianoforte modeling, well...

Which harpsichord do you recommend in Pianoteq? There are several options but I haven't really explored them.
I only tried the free KiViR instruments.

Originally Posted by Frédéric L
Originally Posted by Harpuia
2. Most sampled based VSTs with 8-36 layers that people usually like here, but without interpolations (Ravenscroft, Ivory II ACD, Garritan CFX, Embertone Walker 1955, etc.)
According to the Ivory page :

“Sample Interpolation Technology used for ultrasmooth velocity and note transitions”

Then Synthogy Ivory’s pianos are among the few VST (the only?) which use interpolation and avoid any velocity layers gap.

Indeed Synthogy ACD allows you to reduce the number velocity layers to save memory and interpolates between them. Also its half-pedaling is at least half-decent. Sadly the Steinway they recorded is just not a very great instrument.

Quote
Kontakt enables velocity crossfade (see https://www.adsrsounds.com/kontakt-tutorials/creating-key-and-velocity-crossfades-old/ ) but this doesn’t mean many Kontakt librairies use this feature. I suppose we can even have phase cancellation if samples are not well aligned.

That's exactly the "playability" issue Software users talk about all the time. Low-end digital pianos only provide a single layer and are still pretty playable. However the latter sometimes leads to the "banging the keys" phenomenon when single-layer players switch to an acoustic piano.
Posted By: meghdad Re: Modeling vs Sampling - 09/26/20 02:51 PM
Originally Posted by magicpiano
It's true that the HI-XL has longer samples, but the duration of the sustain is independent of the sample length, because the piano engine can make the looped sustained part long/short almost as you want.
How can I measure the sample length on my piano? How did you do that? Did you connect it to an oscilator and analyze the ouput graph?

I know, in the end what matters the most is "how it sounds to my ears". But I'm just curious about this.
Posted By: MacMacMac Re: Modeling vs Sampling - 09/26/20 03:04 PM
Read the DPBSD thread by dewster. (Go to the "master sticky" thread. There you'll find a link to the DPBSD thread.)
In the first few posts dewster explains his measurement techniques. You might find an answer there.
Originally Posted by meghdad
How can I measure the sample length on my piano? How did you do that? Did you connect it to an oscilator and analyze the ouput graph?

For me, though, the point is moot ...
Originally Posted by meghdad
I know, in the end what matters the most is "how it sounds to my ears". But I'm just curious about this.
Posted By: MartF Re: Modeling vs Sampling - 10/19/20 11:43 PM
I just received an email from NI regarding Cremona Quartet and it mentioned "phase-aligned stereo samples". I don't know if that means on cross-fade, or it's just marketing speak for "we know how to record in stereo". Reminded me of this thread anyhow.
https://www.native-instruments.com/en/products/komplete/cinematic/cremona-quartet/?content=5515

edit: it does say "Phase aligned velocity crossfades on long articulations".
Posted By: VladK Re: Modeling vs Sampling - 10/19/20 11:51 PM
Originally Posted by MartF
I just received an email from NI regarding Cremona Quartet and it mentioned "phase-aligned stereo samples". I don't know if that means on cross-fade, or it's just marketing speak for "we know how to record in stereo". Reminded me of this thread anyhow.
https://www.native-instruments.com/en/products/komplete/cinematic/cremona-quartet/?content=5515

edit: it does say "Phase aligned velocity crossfades on long articulations".

phase-aligned probably mean that they apply calculated delays to samples recorded with different pairs of stereo mics, i.e. ambience mics sound with, say, 20ms delay to close ones. This will result in more pleasant echo/ambience/reverb thanks to proper transition between, and mix of, close and ambient sample velocitiess.
© Piano World Piano & Digital Piano Forums