Piano World Home Page Forums Our Most Popular Forums Digital Pianos - Electronic Pianos - Synths & Keyboards Pianoteq, Samples and Phil Best

Hi Everyone,

I have to say Ive been puzzled and pained lately by the intense, passionate and sharply critical narrative that sometimes blows up over the desirability or otherwise of sample based piano software verses pianoteq. It’s a debate that seems to polarise views even when this is not the initial ntention.

So, I’m going to attempt to advance a theory based on my own impressions of why this might be happening. To begin, I do love pianoteq and playing the pianos gives me immense pleasure and as a struggling low to intermediate student I need all the help I can get. In addition I own and play the sample based Vintage D which I love and the CFX in my Clavinova 709.

So here is my experience of pianoteq. If I play using my iLoud Micro Monitor speakers hooked up to my Apple Mac Pro, the pianos seem fine, even wonderful at times. BUT if I sit down fresh and hook up my Sennhieser 598 headphones to the apple and Pianoteq, the sound is problematic. Some of the pianos and I’m thinking Steinway D among others can sound overly metallic and somehow a bit electronic in timbre. However, after a couple of minutes of adjustment this impression fades and the pianos sound lovely again. What on earth is going on here? I have to say this is not the case with the Vintage D. If you think the description of my initial headphone experience sounds similar to those criticisms levelled at Pianoteq you’d be right.

So what might be going on here? Well, one thought I’ve had is that some people are particularly sensitive to the Pianoteq electronic signature whilst other are not or much less so. I don’t even think it’s about taste exactly but something possibly wired in that makes one person hear and focus on this midi like aspect of the sound. For me, if Pianoteq remained sounding electronic, I too would be buying and advocating samples. Most of the time I’m not sensitive to it and so ignorance is truly bliss.

Where then does Phil Best come into the picture? As many of you may know Phil is a great supporter of Pianoteq and posts videos of his wonderful improvised piano playing. I know, I am an envious fan. His latest video ‘how to get the best sound out of Pianoteq’ explores his views on this matter. It can be viewed from the link below.

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

I hope the sample guys will forgive Phils reference to the unresponsiveness he finds in sampled pianos. This is not what I’m concerned with or something I am aware of or agree with at my level of expertise. I do think however that the way one plays pianoteq may affect the timbre issues and under certain conditions aggravate it. I have a feeling Phil could make even a bad piano sound half decent but it does raise interesting points about touch and technique.

Anyway, what do you think? Does this make any sense or am I barking up the wrong tree.It wouldn’t be the first time😄

Paul

I think you've presented some good points without emotions. And also I think it is probably good that there were sometimes heated debates with a lot of emotions because that's ultimately serving a purpose: to show that not all is good, on both sides.

Once we overcome those emotions and personal stuff, we could concentrate and see what are cons and pros on each side.

I will describe my work with the Pianoteq beta team and according to the NDA I signed at the time it has expired so I may uncover my participation (which I have already done in other threads and I guess Pianoteq guys will be OK with that since I am not violating the NDA).

When I participated in the Grotrian testing, I devised a way of comparing my favorite piano at the time: the Vintage D and the Grotrian releases. I had Vintage D and Pianoteq both loaded in a DAW listening to different MIDI channels. And on my Kawai MP6 I switched between two presets I created that would use one or the other MIDI channel. Thus I could immediately switch between playing Vintage D and Pianoteq and pinpoint particular phrases, note sequences, etc where it's really apparent how Vintage D sounds and how Pianoteq sounds. Using this direct comparison approach to me there was apparent problem with Pianoteq when playing lines, like simple scales with short attacks, especially in the mids and mid-highs. I think Philippe and the team liked that approach because they were able to improve what I have reported and besides I was sending them both audio files as well as the MIDI file which I guess was very easy for them to reproduce and compare with newer configurations.

I think Phil Best is wrong about the unresponsiveness of sampled pianos, unless he just means the majority of sampled pianos. There are some, though very few, sampled pianos that are as responsive as modelled ones.

There are a lot of misconceptions about sampled pianos (and I guess about modeled ones but since I'm on the sampled side, I can't comment on that). For instance, I've seen many people commenting on how a sampled piano has e.g. 6 sample layers and so you're always hearing only 6 different velocities That has never been true. And there's even sample interpolation as Dewster has demonstrated with some of his measurements of pianos where from the spectral diagram a spectral content steps were not seen to suggest sample switches, it was a smooth progression 1-127 both as sound volume and as timbre gradation. That has been the cause for many people simply assuming sampled pianos are not "playable" or not "responsive" enough.

@pmh:
I always thought Pianoteq lacks something. Recently it has improved a lot but there is always some perception of "digital" that I feel in the generated sound (especially in the middle range of the keyboard). High quality sample-based VST maybe are less versatile but are superior in sound quality, in my opinion. I never listened yet to a pianoteq settings that mimics a vintage Steinway Model D with the same quality and level of detail as a VST like "Ivory 2 American Concert D" does. At the first you feel it's there, but after some playing... The ear gets used to it and you are able to feel that digital "something" that bothers you.

But who do you think wins between Pianoteq and the sound engine of your Clavinova?

I think modeled piano synthesis is great and, as a developer and piano enthusiast, in my free time I'm having fun in trying to develop a modeled piano engine that could add that "something" that Pianoteq doesn't do yet. It's not easy at all and it requires much time for a single man to do research and experimentations, so I don't know if I will ever finish this project. But who knows... For now I made an experimental VST that can play 2 octaves of a Steinway-like piano sound (after 1.5 year of work) and I think it sounds very nice.

magicpiano, that's really interesting! Can you uncover more details about your engine? What kind of modeling have you implemented, etc. Of course, if that's not a secret Can you share some examples?

But not all the VST do interpolation, so in some of them you can hear clearly the sudden passage from a layer to the next (in "Piano in Blue", for example, I had this perception). Honestly, I think that VST with 16 or more layers don't interpolate at all. With 20 or more layers I think the human ear cannot perceive a clear passage from a layer to the next.
DP engines have to do interpolation because they use very few layers (from 3 to 6, I suppose).

I agree with CyberGene, this is really interesting. I want to do the same thing, but it will probably be around a year or so from now, before I will have the time to work on it for real. I hope you can develop something really useful. Whatever you do, I suggest you give the modelled sound more sustain than Pianoteq, whose sounds die out far too early (when playing a note and holding the key down).

If you want some friendly criticism of the sound, post some samples for us to hear. Maybe we can provide input for improvements.

Best of luck with the project!

All I can say is that it's a mix of techniques. At first I create the base waveform of each note and then use various algorithms to add beating, dispersion, and other effects, like the key release, in real-time. I create a "mono" note and then apply some techniques to separate the sound in left and right channel. One thing I'm unable to create in a realistic way with modeling is the damn hammer sound, so it's the only sample I add at the very start of the attack. The rest is algorithmically generated.
Maybe sometimes I will make a youtube video with me playing some notes with it.

I think when it comes down to Pianoteq versus Sampled sounds it all depends upon what is the purpose you have for using either one of them. If it was for critical listening as in recording yourself for playback some VSTs will definitely produce a better "sound" because they are digital recordings of actual piano sounds. If it was for critical practice or playing I would go with Pianoteq's modeled pianos because behave more closely like a real piano. There is more sensitivity to your input in mine and many experienced pianists experience with Pianoteq. So in my opinion Pianoteq is the best compromise between sound quality and performance so it is my go to application.

Yes I've had a clear stepwise layering when listening to certain digital pianos rather than a smooth linear response to my inputs. That's exactly the lack of sensitivity I'm referring to that pushes me towards Pianoteq. It's response is more linear.

I can certainly hear no sample timbre switch with Garritan CFX. Maybe it's not even interpolated but having enough sample layers and only changing the volume is good enough. Remember, the piano is not a synth that changes drastically its timbre from velocity 10 to velocity 11. IMO having around 20 layers is absolutely enough. If you've played digital pianos from the past, then yes, those may have 3 layers with no interpolation and so obvious timbre switching is present. Which isn't a reason to dismiss all sampled pianos, especially huge libraries from the past few years, based on old Clavinovas and other outdated instruments.

I think user "EvilDragon" is a beta tester or consultant at PianoTeq as he seems to have plenty of inside information.

I think EvilDragon said PianoTeq also uses samples for the early attack.

I think EvilDragon also said that PianoTeq will record a piano and compare the samples to the model output and tweak.

I think CyberGene linked some of the PianoTeq patents and whitepapers (but at PianoWorld forum).

Was there a Russian guy who built a beta modeled piano a few years ago?

Of course I can't find those posts and I might be mis-attributing.

I wonder whether this Pianoteq "Evil Dragon" is the same "Evil Dragon" who offers expert advice on Native Instruments' Kontakt board ?

He is truly an expert at using Kontakt scripting. There are ways to learn, and he has amply done so.

But I don't know how he (or anyone) can be expert at the innerds of any piano software, save for those who work on it. And those folks are likely obliged to be silent on the matter.

It's the same person. He knows a lot about Pianoteq, but Philippe who created it has often discussed bits and pieces about its inner workings. I'd imagine a Pianoteq nerd would have collected a lot of information over the years.

I think it is best for each individual to just use your ears and your touch and decide for yourself. Just understand modeled sounds are an option out there and may be the right choice for certain pianists who need a certain level of realism in their digital piano.

Hi Cybergene,
Thanks for the inside info on your Grotrian project. Most interesting. What I have tried to do is find a reason for the marked disparity between peoples impressions of these piano sounds in order to create more understanding about the perceived differences. Hopefully this might lessen the sound and fury and allow for a more tolerant, balanced and inclusive debate.

Hi Magicpiano,
Personally I’m not aware of missing anything in the pianoteq sound. The more I listen, the appealing they sound. In comparison with the Yammy CFX and Bosendorfer, I do prefer the Vintage D and my pianoteq pianos. These include the Bechstein, Bluethner and Steinway. Having said this I recently compared the new Yamaha N1X with the Kawai Novas. I liked the Novas but loved the N1X . The touch is amazing and it has fantastic dynamic performance. I also tried a C1X grand and the touch seemed identical to the digital N1X brother. I also thought the binaural CFX through headphones was awesome. If only I’d waited four years.

Hi Jethro,
I simply enjoy playing the pianoteq stuff. In one mood it hits the spot, in another its the vintage D. I think it’s great to have so many different choices.

Paul

Anything that depends on the competition being somewhat inadequate to be successful won't last in the long term. I agree that some VST sample libraries are bad, but these aren't the competition for the forum's most loved piano software.

Hardware synthesis surpassed computer software and is currently the state of the art for digital pianos. The term "sample" is highly inappropriate for what is happening in modern HW synthesizers.

For example, Ivory II American Concert D offers high quality samples with many layers including una-corda samples, but its software synthesizer is completely outdated. Low polyphony, last decade resonance modeling and very basic half-pedaling. Yamaha's VRM and Roland's SuperNATURAL are far superior. And while Pianoteqs "modeling" capabilities might be farther than other VSTs it still doesn't sound as piano-like as hardware does.

Not saying Ivory ACD is the best, i also find the resonance very dry but you can increase the voices to 1000. It's usually set to automatically switch to 1000 when you're rendering your recorded track but you can do it while playing as well even though it may tax your resources a bit.

I like to see more videos like this. Phil Best is such a pro pianist that I can listen to him improvising for the whole day. It doesn't matter which piano you like but Phil did convey his understanding of good pianos.

One note is stronger than 1,000 words. It's better to see someone explains their opinion by playing the pianos themselves.