Sample pianos - R.I.P.? (Page 1 of 3)

I agree with Beto-music with this. In fact you can hear this "lack of resonances" in many sample piano demos. But back to business. What's the main and most important thing? In my opinion it is the basic sound (resonances come second) and what I haven't had yet from PTQ is this modern studio sound I described earlier. Taking digital thing to a Pop/Rock/Jazz gig what you need is good basic sound. And if you don't get that it doesn't matter how much does it resonate.

I don't know what is YC5 with Studio Close Mic but to my ears it's far from sound of playing real Yamaha Grand. (And I have said this many times elsewhere: PTQ's weakest part is around Middle C and going lower from that. In higher registers PTQ is doing well...)

P.S: I'm sorry if I sound critical but sometimes I feel that somebody in this forum must be a bad guy - PTQ is very good and promising thing but far from perfect (same is with sample pianos of course!)

Sorry but just at moment I cannot give you very spesific example what I am after for. (And I doubt that individual notes doesn't tell the whole thing).

If you hadn't done it yet go to synthogy's  page and listen to "Rock Ballad" example of C7 to get some idea.

http://www.synthogy.com/demos/grandpiano.html

Some of the people here might say that it doesn't sound good or have a "soul". Reason I give you this example is that in my opinion this part - middle C and lower from that - is the most difficult for PTQ if it tries to emulate this type of very clear studio piano sound. If you wish, pick up the most close preset of PTQ and play same chords (C, F etc.) as the synthogy demo. Does it sound same? I can get something like that from PTQ but obviously something is missing and for me it's sounding little artificial.

As I said earlier in the upper registers PTQ is doing well.

I hear it and know what you mean, but in order to have a valid comparison, we need to have the same midi file converted to wave by Synthogy and Pianoteq.

I'm not making excuses, but the lower notes are harder to create with physmod.  I don't think it's any more difficult to render mathematically, it's that most of our computers can't handle the multitude of frequencies from the low bass.  Copper wound bass strings emit a lot of odd frequencies and it isn't unusual that these require re-working or replacement (my Yamaha did).

Could you post a midi file that you've rendered in Synthogy to wave, that we could attempt to emulate in Pianoteq?  You'd have to post the mp3 too of course.

Glenn

Ecaroh,

You're absolutely right: the sound you're referring to is, for the moment anyway, completely out of Pianoteq's reach. No need to compare or to analyze or to investigate anything, it simply is a sound that can't be coaxed out of Pianoteq, not for the time being. Pianoteq covers a whole lot of ground, pianosound-wise, but obviously it can't cover all the ground. Well, not yet anyway. And until it does, there remain quite a few pianotimbres (not just the one you're inquiring about here) which are very difficult, if not impossible, to create with Pianoteq.

So, yes: Pianoteq, today, doesn't really have anything in its arsenal of instruments that suits that Corea-type of playing which you mentioned. You'd need an instrument with a dry, crisp sort of immediacy and clarity to it, and that's precisely a timbre that Pianoteq still doesn't have.
But I'm confident that it will happen. No doubt about it. That's one of the great things about Pianoteq, you see: give Philippe and the rest of the Modartt team a challenge like this and they won't sleep until they've cracked its secret. So, just sit back and wait. What isn't today, I'm sure of it, will be tomorrow. That's Pianoteq for you.

_

Samples are Edison's technology - record a sound and play it back.  If you want to hear Ella Fitzgerald sing a particular song, this is still a good method.  If you want to hear her sing another song, you'd better have a recording of that too because you can't take samples of  her voice and use them in the second song.  Admittedly, the piano is far easier to emulate than a human voice or a violin/guitar/sax/trumpet/etc, but the principle is the same.  A sample after all is just a sampling (only a small part of the whole).

I won't say samples don't have a wonderful sound, they do, but I play the piano, and I find Pianoteq responds extremely well on the keyboard.

There are always going to be trade-offs, and I'd rather have something that works well than something that sounds great but doesn't work when I play it.

Glenn

Well, you have to remamber that Modartt is limited by consumers computer power.
My bet is that they can do all emulations of a great sound including some aspects that is not present yet. But it would take more computer power and it could turn the software too heavy to average users.

Perhaps V-Piano gave different priorities than pianoteq, while choosing what to emulate better, tone or richness in ressonance effects.

The challenge is to create a algorithm that is at same time precise and lightweight enough to generate a great ultra realistic  sound, and still have a reasonavle CPU consume.

We can be sure about one thing, soon the modelled pianos will really kick the sampled pianos. It's perhaps 60% matter of computer get more powerfull processing capability, and 40% matter of algorithm development .

That's "Symphonic Choirs".

And the words are rather vague with most any massive choir.

Piet:

The mathematicians need to get together perhaps?

Glenn

Yes, I suppose that is true - thanks.

Greg.

Let's not give up so fast. We may be able to get this sound. We can control the spectrum profile, and lowering the Strike point on the C3 creates a more brittle attack. Sounds as though there may be a mic at the rear of the piano, too?

I’m sorry, Jake, but I’m dead certain that we won’t. That sound simply is not present among Pianoteq’s current algorithms. It’s not even gestating there at the moment. It is fundamentally different from anything that is, at present, available in Pianoteq and no amount of spectrum profiling, no amount of parameter adjusting can change that. In order to create the ‘Ecorah sound’ (let’s call it that, shall we?), Pianoteq needs a completely new model, added to its existing models. It needs a whole new set of additional algorithms.
There are just too many characteristics (and all very important ones) in that Ecorah sound which don’t appear yet on Pianoteq’ current pallette of ‘pianotimbre ingredients’: the dry attack, the immediacy of the sound, the close-up clarity, the organic openness, the detailed definition of the tone ....
What you’re suggesting - with admirable optimism and confidence, I’ll give you that - would be comparable to trying and create a brash, brilliant Dizzy Gillespie-type trumpet sound out of the smooth velvety Miles Davis flügelhorn sound. Just can’t be done, can it? (Well, not convincingly anyway.) Not with all the EQ’ing and digital processing in the world.
I really think that we’ll have to wait a little while longer before Pianoteq will include an Ecorah piano. But, again, I’m positive it will happen.

_

Actually, I think you hit on a great point! Thanks to Pianoteq's playability, I really don't need the greatest sounding real-time model, at least for recording. So I suppose they could delve a lot deeper in terms of "rendering" and such after the fact. I'm not sure how that would work, but it sounds promising. And of course eventually computers will catch up and be able to handle the more complex algorithms in real time.

The problem with trying to "fix" this with the spectrum profile, as I see it, is that one is covering the frequency domain but not the time-domain.  I know from bitter experience with organ samples that an unfavourable attack profile can't be compensated through equalization.  The situation with the piano is  broadly similar - the detail in the initial attack is the problem.  This isn't going to be fixed by any kind of equalization; the model would need to be altered. Who's to say that the changes couldn't be made without extra CPU cycles?

This may also come back to the "assumptions" in the model which Modartt have mentioned in the past; model elements which are not currently calculated in real-time to save CPU cycles. 

If Modartt know what is "missing" from the current PTQ sound, WHY it is missing and what extra computations would be required to rectify it, then what is the resource cost:benefit ratio like and how many hardware choices would be excluded from running it?

My guess would be that dual-core machines (especially laptops) would be out and maybe the base limit would be a Mac Pro/dual Xeon setup.  We're then into the same territory as Hauptwerk and edging towards a very expensive hobby.  A happy medium to me at present would be to assume that for top performance, a quad core CPU was available.  Like with Hauptwerk 3, a musician could use a lower-spec machine *but* advanced features would have to be turned off for performance to be acceptable.  It should however always be possible to tune the settings to get acceptable performance on any contemporary hardware.

Of course the situation might be that the current model doesn't include the solution to the sonic problem - as Piet suggests - therefore offline use/more horsepower isn't currently an option.  If the model doesn't cover it, then is it even possible to say what the computational requirement might be?

Best//Neil

The metallic quality that I hear is not only in the initial attack - it lasts well into the sustain. A steely, wiry quality, that evolves in a pleasant way. (similar to the "phaser" effect) I know that unison detuning makes the sound evolve, but Pianoteq is not producing the same detail yet, IMHO. From the V-Piano demos, I agree that it's closer. (but still not there yet)

Greg.

Just to refresh some consideration that was already discussed in this forum.

-V-Piano despite of have many processor, have no such big deal of processing power in end, cause the processors are not very fast.
I bet V-piano cut costs, of CPU consume, by rendering a inferior quality resonance effect than pianoteq, and invest this saved power in tone rendering.

-Modartt original intention about pianoteq was to create a piano software that people could load and run from a laptop, and present all main living ressonances an colors of a real piano.

In the beggining of this forum I had mentioned at least two times, about a heavy processing to more elaborated piano timbre for export audio.

Maybe a alternative would be a version with options to able or disable some computations of piano effects.  People with a "turbinated PCs" would sellect a heavier mathematical computation for sound generation.

This would require time effort, and the market must be considered. How many people, who enjoy realistic Grand piano sound, have "turbinate PCs" ????   Unfortunately not much...

I had a quick look at that 'Rock Ballad' preset in Ivory and it has chorus added to it. Which makes a big difference (a.o. the 'phaser' effect that Skip noticed).

But it seems to me that we're talking about two entirely different sounds here. That 'C7 Rock Ballad' sound is certainly NOT a sound that I would associate with Chick Corea to be honest: it's a fairly dirty, uneven sound with a gritty, slightly worn-out character, and I think much more at home on a Tom Waits, Little Feat or Lynyrd Skynyrd album than on a typical Corea record.
When I think of Corea, I think of a really clean, very dynamic and lively sound, with lots of detail and a crystal-clear definition. Nothing at all like that 'Rock Ballad' piano in fact.

And that (the Corea sound I just described) is the sound I was thinking of when I said what I said in my previous posts, about Pianoteq not being able yet to deliver it.
The 'Rock Ballad' sound is, at the moment, out of Pianoteq's reach too, I believe (there are organic, 'non linear' things going on in that timbre which I can't see Pianoteq doing just yet), but trying to emulate it seems certainly a much more realistic idea than trying to recreate the Corea sound.

_

Hi Jake,

Here's a pretty good example, a few fragments from Corea's "Acoustic Band" album (with Patitucci & Weckl):
http://users.telenet.be/deridderpiet.be/Corea01.mp3

It's not so much the timbre itself that I like, you know, it's most of all the unbelievably snappy character of this pianosound. Just listen to those extremely crisp staccato chords and the liveliness of the phrases ... that's the sort of thing I'm talking about: snappyness, clarity, directness, strong articulation, etc. ...

_

your comment about "hybrid technology" - it seems intuitive to me that you could start with a recorded sample note of a fine piano and then model all the things that happen from there.  but maybe the changes that occur in timbre at different velocities are best modelled completely vs "altering" a sample.  i just don't believe purely sampled pianos will ever get it done.  i can see them in a mix where the overtones and resonance get crowded out but i can't ever imagine them standing up in a solo recording.  i can just see somebody putting out a dozen blu-ray disk multi-nanobyte sample program where your lights start dimming when you play more than two notes.... Lord help us..

Skip,

The ArtVista is famous for its singing quality (and you're right: it's not a chorus effect). It sings so much in fact, that quite a few people consider it to be badly tuned. But it is also a quality that many people seem to like very much: the Vista is among the most popular sampled piano's. There were a few passionate discussions though, on several forums, about the Vista's idiosyncratic tuning during the first months after its release.

_

Ooppss, forgot to mention that Pro Mega 3 came after Kornel Mezzo scripts.
Pro Mega 3 had a algorithm to create seamless velocity changes, and a processor to try to emulate a lot of ressonance effects.

But no one talk about Promega 3 today.

"The PTQ piano model, with all features enabled, will achieve the maximum required polyphony for the Grieg Piano Concerto, on the cheapest variant of [Intel's CPU line with the highest available core count]"

(I.e. for Dec 2009, it would achieve this on a  Quad-core Q8400 system.)

I'd like to see this as a design brief for future builds of Pianoteq.

This is NOT a high-end system and could be built for not much more than a decent masterkeyboard.  It is, however, vastly more capable than the current requirements of PTQ.  I run PTQ on a Pentium M I built 4 years ago.  I can achieve polyphony of 60 before problems, and under most circumstances, it's enough.  I'm certainly not going to complain that my PC isn't up to the job, as I know it isn't.

My view is that tuning for speed should be possible, with lower-spec computers, but making the PTQ model run properly - with all bells and whistles engaged - on:

Celerons
Atoms
Single-cores
Core Duo
Pentium Dual-Core

...shouldn't be in the design brief, otherwise progress won't be made.

I admire (I think) people attempting to run PTQ on a netbook, but ultimately this is counter-productive.  Most decent masterkeyboards suitable for PTQ cost more than a netbook.  A decent soundcard is (upwards of) half the cost of a netbook.

That is telling.  I hold the same dim view about ASIO4ALL.  Yes, it's a great tool, but the better option would be to buy a decent external soundcard, something like an Edirol UA-25X.

Martin Dyde has the right idea about performance envelopes, with Hauptwerk 2/3.  Do the job properly and don't spare the horses.  The result is that soon, Salisbury Cathedral Organ will be available in the living room of anyone who can a) afford the samples and b) a Mac Pro to run it.  Can't afford a Mac Pro? (this includes me)  Tough.  Tant pis! 

Rethink your expectations.  I did, and that's why recreating a fine grand piano in my home is now my prime goal, not accurate performance of Elgar's Organ Concerto.

If Martin had continued to develop Hauptwerk so it ran impeccably on single-core machines, the wealth of historic sample projects available would never have become reality. The software wouldn't have broken away from the shareware, hobbyist market it was originally designed for.  No-one would have taken it seriously.

Do you want a more realistic and accurate digital piano, used and abused in studios all over the world, or do you want something that plays well on a PC you can get for free if you (for example) sign up for 3G wireless service?

Fractal Audio's AxeFX has been mentioned here on occasion.  Yes, it's expensive.  But it's a compact 19" rack which costs less than one bulky and fragile valve head and can emulate just about any hardware you want.  Surprise surprise it sells like hot cakes.

As was once uttered in Field of Dreams, "If [they] build it, people will come."  A more advanced PTQ model would only attract more custom and a higher profile. Netbook users' noses would only be out of joint if it wasn't possible to scale the model improvements back to manage performance.

As an aside, my missus recently tortured me with the DVD of Take That's "The Circus" live show.  For several numbers, Gary Barlow was playing a Roland V Piano, as part of a busy mix.  It sounded "OK" but didn't blow my skirt up.  It certainly wasn't a patch on the Yamaha grand he used for exposed, solo piano later on.  The upper register of the V-Piano sounded extremely artificial.  The rest sounded just like a regular digital piano.  A major stadium tour for V Piano is a lot of exposure though, if anyone in the audience was looking...

(The pro version has some pretty astonishing abilities. I'm still just scratching its surface. The more I try to match notes to the exact sound of recorded notes, the better the results, for me.)

Glenn,

Here's some of that Ivory C7 'Rock Ballad' (both the audio and the midi-file):
audio: http://users.telenet.be/deridderpiet.be...Ballad.mp3
midi: http://users.telenet.be/deridderpiet.be...Ballad.mid

_

Thanks to all for your very interesting comments. I fully agree that the potential of innovation and improvement of Pianoteq is great and we are currently working hard on new models. It is a very challenging task, and we cannot tell how long it will take.

Meanwhile, I invite you to explore the Pro version. Even if it does not allow to obtain any kind of piano sound as observed by Piet, it is certainly capable of producing many interesting and maybe unexpected things. It’s the beginning of a new story!