Resonance: Why Sampled Pianos NEVER Sound as Good as Real Pianos

For those who keep commenting that I am wrong because some sampled pianos now use physical modelling to create resonance, you seem to be missing the point - and at the same time even reinforcing it - since those pianos are therefore no longer just sampled, but hybrids, and it's the physical modelling that sorts out the resonance. And that means you actually agree with me, but simply didn't think through the premise of my argument, nor that of your own.
I do apologise for sounding a bit absolutist in this video, but resonance is perhaps the one thing that is the most difficult to get quite right on sampled pianos, because of the challenges in recording the near-infinite variability that exists in the natural behaviour of a complex system of resonances.
Some sampled pianos do include recordings of resonant strings and the programming needed for them to sound in the correct conditions, like The Giant from NI, Korg Kronos, and some other ones. While a sample cannot accurately cover all possibilities, you can get a lot of mileage out of just the octaves and fifths, for instance.
Modelled pianos like Pianoteq and Roland V, on the other hand, can calculate sympathetic resonance precisely because it simulates the entire system, all its components, and the exact physics of the keys being struck, which is why they can consistently sound even more like real pianos - or even unreal ones like the super-resonant one I demonstrate in this video.
COFFEE and PATREON:
/ hakonsoreide
www.buymeacoff...
ko-fi.com/hako...
If my channel ever inspires you to buy any audio gear, please consider getting it from THOMANN. If you visit them through this link, it makes me a little bit at no extra cost to you:
www.thomann.de...
Also consider buying a piece of music, or an art print to support my work:
MUSIC: hakon1.bandcam...
ART: fineartamerica...
Please also visit Poetry! My channel dedicated to all things lyrical:
• Video
#Pianoteq #Yamaha #Piano

[@cmtwei9605]

3:55 onwards: the resonance of the piano after a key is struck comes from the sounboard, the sides etc., also from the duplex section, that is the front and rear parts of a string beyond the capo d'astro bar on the front and holding pin at the rear that is not affected by the damper. The front and rear lengths have been shown to vibrate outside the expected harmonic frequencies in studies. Also, the very top one and half octave in the treble strings do not have dampers. Modelling algorithm tries to simulate what happens when other keys are struck and others strings vibrate. It can only be a gross estimate, hence modelling. In real situations the interaction must be so complex with some randomness so it cannot be precisely calculated. Air humidity affects the strings and dampers so the same piano can sound different on different days.

[@hakonsoreide]

Yes, piano resonance in real life is a hugely complex thing that would, as you say, vary depending on changing outside variables. All physical modelling instruments, as far as I know, mathematically simplifies resonators down to two-dimensional objects as that makes it manageable to calculate without adding too much latency or processing requirements. Because even the tiniest variables make such a difference to the resonance, this is also why two pianos, made in exactly the same way, by the same people, from the same materials won't sound the same even when right next to each other in the same environment. It's simply impossible to make them identical. I like how in Pianoteq you can make some small adjustments to the parameters and effectively make your own piano that will sound differerent to everyone else's. Another aspect of acoustic instruments compared to digital ones, of course, is the experience of hearing them in a three-dimensional space as opposed to a recording when the sound of an acoustic one would emanate not just from two points but from the entire resonating body of the instrument, with different frequencies coming off stronger from certain parts of it, and the tiny differences in direction and timing of when they reach your ears affecting what it sounds like. A good stereo recording, or stereo recording emulation as in Pianoteq, only creates an illusion of a three-dimensional physical space, but it still won't be able to entirely capture it.

[@gearogear]

Thank you very much for articulating what was in the air but could not be expressed in words.

[@hakonsoreide]

Thanks for your comment. I was quite fascinated when I realised this as I had heard the difference ever since I played my first electronic piano many years ago, but didn't know why or what it was yet. Thankfully, more people are aware, and it doesn't take a lot of resonance sampling or modelling added to a sampled piano to fool the air into thinking it sounds more real - and that is something that does happen these days. I'll still use Pianoteq, though. I just love how you can make pianos that don't exist, and how it doesn't take up a huge chunk of storage space on my computer.

[@Musicwithoutthought]

Thanks for the video! Nicely done! A couple of thoughts. First, some sampled piano libraries have sympathetic resonance (e.g, Ivory, some of the NI pianos). I don’t know how they compare to Pianoteq. Second, reverb can be used on piano to place the piano in a room rather than create sympathetic resonance. I view them as separate effects. Third, the advantage, in my view, of a real piano is the feel and vibrational feedback from a wood instrument. A recording of a “real” piano is a digital or analog representation of piano. Pianoteq is my go to piano mainly because of the feel - it is so responsive. They keep improving the sound.

[@hakonsoreide]

Thanks for watching and commenting. Some excellent points. I just got a tip that The Giant from NI has sympathetic resonance. It is off by default, but it sounds very close to the real thing or a modelled piano when switched on. Another thing, that is actually hard to replicate with any acoustic instrument is how their resonance is more three-dimensional than any stereo rendition, sampled or modelled, can accurately convey. You are of course right that reverb (and eq and filtering) also is a tool for perceptual placement in a space, so my rather blunt statement that you don't really need it for pianos is of course a little hyperbolic. I started with Pianoteq, and after that I haven't found many other piano sounds I would choose above it, but I kept a few of the NI ones with interesting characters, The Giant being one of them, but I still feel I get more mileage out of adjusting parameters in Pianoteq, and the default response feels more like a real piano to me.

[@Musicwithoutthought]

The Giant is a unique, one of a kind instrument from Klavins (as is Una Corda) - highly recommended for the unique sound. But Pianoteq is my go to piano. The newest version has a morphing feature which is amazing.

[@hakonsoreide]

Oh yes, I saw the announcement video for the Pianoteq morphing. Seems like a cool feature. I will upgrade at some point, but still keep Pianoteq 4 installed for its resonance quality.

[@Oshvedamusic]

Oh my god. This is the BEST video on (Modeled VS Sampled) Pianos. I watched 10 videos but none of them showed facts about resonance of pressing a note when holding other keys down.

[@hakonsoreide]

Thanks. Yes, I was wondering for a while why sampled pianos sounded so lifeless to me compared to the real thing (and to modelled ones) until I heard about that particular thing, and it all made more sense. Self resonance both in the open strings and in the instrument itself is such a huge part of a piano's sound, and it's caused by the interaction of _all_ the sounds vibrating at the same time in the instrument.

[@Merlijny2k]

When you read about DP's and view DP demo's they all write and talk about 'dynamic range' being better on modelled piano's and so on, but no one makes it simple understandable and audible like you do. Thanks for the great video!

[@hakonsoreide]

Thanks. Dynamic range in itself isn't the most defining factor, but dynamic resolution certainly can be. A sampled piano, even the ridiculously huge ones that have a velocity layer for each of the 7-bit MIDI values of 0-127 still can't beat the 16,384 discrete values that Pianoteq can respond to. And for any sampled piano with fewer layers (which is most), the problem isn't dynamic range, but the fact that within a velocity layer, the timbre is unaffected by the change in dynamics when on a real piano the timbre continuously changes through the dynamic range, as well as through interaction with the other physical forces at work simultaneously, such as other strings vibrating at the same time, affecting each other and the soundboard at the same time, something you simply cannot sample for all possible combinations of parameters.

[@victornoagbodji]

Thanks for making this video 🙏😊 This helps me understand why music I am currently working on sounds the way it does. It happens when I play 4 or more notes on my piano VST. I can hear a nasty resonance. Now, I use a sampled VST but they must be doing more that is resulting in the artifact I hear. I also thought of a musical solution, to borrow a phrase by another KZbinr. I will duplicate the piano part and have separate low and high parts.

[@hakonsoreide]

You're welcome, and thanks for your comment. Luckily more and more piano virtual instruments and keyboards are using modelled or hybrid sound generation now, the hybrid approach using modelling specifically for the resonance and sampling for the played keys, so they will keep sounding better. At the same time, of course, Pianoteq, which is now 3 major versions further along than the one I demonstrate in the video, is sounding even better, and I still love how it allows me to basically "build" my own piano by defining its resonance, damping, string length, and other properties of how sound is generated and resonates through it.

[@crisoliveira2644]

In 5:23 you state most sampled pianos do not sample sympathetic resonance. That means you know there are exceptions. I just think never is too strong a word. Not that there may be a sampled piano that sounds as good as a real one, but there are sampled pianos with resonance. I also think that feature is a must in software pianos by now. When you pay to have gigabytes of sampled piano, you'll want to have all the features.

[@hakonsoreide]

There are a few sampled pianos that do resonance. They still don't sound as good as real or modelled pianos, so I still think my semi-clickbaity "never" is justified, but it is getting close. Since it is super inefficient to sample every single velocity level together with every possible combination of other keys that might be held down, they use a hybrid sound generation with sampled played notes and modelled resonance. At the same time, of course, Pianoteq keeps improving their algorithms all the time and are well ahead. I also like that I can use it to make my own piano sound rather than having to use anyone else's.

[@crisoliveira2644]

@@hakonsoreide Funny that there are people who won't touch Pianoteq even with a stick. I too used to dislike its sound, but as I got better, I started to enjoy it. But it seems it's a love and hate situation. Kinda like the blue/white and gold dress phenomenon, you see it in one way or the other, there's no in-between and you can't imagine what other people are seeing in it in order for their opinion to be the opposite of yours. By the way, some people realised you can set Pianoteq to zero velocity, play it alongside sampled pianos and have both the sampled timbre and the resonance.

[@Justin_collier]

I really think if you get the right keyboard like a kawai or nord you can get the actual sound some grand pianos go out of tune and i feel at times that my kawai can sound better when a grand is not in tune

[@hakonsoreide]

If you want a piano to be out of tune, that is of course another good reason to use modelled pianos. You can make it go pleasingly out of tune just by adjusting a slider, or by adjusting each note individually, for example, in Pianoteq Pro.

[@Justin_collier]

@@hakonsoreide I myself prefer digital over accoustic for reasons. I actually have my own belief that if you tweak the parameters in virtual tech, I put my kawai on the dynamic setting than set the touch response and damper resonance on half, then EQ it , to me i feel i can get about the sound of the actual thing, but When I go to my church, i play in a band, all those effects in virtual tech get toned out by the time instruments play just as they do get toned out on a real acoustic.

[@finnurjuliusson9635]

Have you tried screaming into a real piano? I am quite sure it does not work on Pianoteq, but they might code it ;D. I understand why you like the longer reasonances of Pianoteq version 4 . Beautiful sound. I myself have version 7 which is lovely for normal piano work but lacks this great feature. At least in the cheapest version(stage I think).

[@hakonsoreide]

Thanks. Hehe. Good point. External audio input for a modelled piano? I am sure that never even crossed their minds, but that would have been cool. I started using Pianoteq on version 3, and it was on 4 I wanted that super resonant sounds. I had to reinstall it after I realised 5 didn't do it anymore, though. I just love how the resonance teeters on the verge of self-oscillation, a sound I also like finding when playing synth filters. I will probably upgrade Pianoteq at some point, but I will always keep 4 around for that particular sound.

[@finnurjuliusson9635]

@@hakonsoreide Yeah for example guitar into your Pianoteq version 4 sound I bet would sound awesome. Btw I have a red phone in my studio too. ;D I must admit it is not connected.

[@hakonsoreide]

Mine is not connected either. I've even removed the cable as it just gets in the way, but I might jig it so I can use it as a microphone some day. It's really hard to get original Norwegian phones from the 60's and 70's as they all were the official property of the national telecom company and had to be handed back when everyone was upgraded to digital phone centrals. Presumably most of them were just dumped when they were no longer needed. When I saw this one for sale, I just had to buy it. I was really looking for the grey ones everyone had at home, but a red one is the next best thing, and also it stands out more because of its colour.

[@finnurjuliusson9635]

​@@hakonsoreide I am from Iceland and we have almost the exact same telephone history. I got mine while clearing out a bankrupt kiosk. My friend connected a grey one as a mic, back in the day. But he had to use the speaker as mic.It is easier as the mic is a condenser and needs extra circuitry.

[@hakonsoreide]

Yes, you may even have had the exact same telephones in Iceland. I know Elektrisk Bureau, the Oslo company that made the ones we had in Norway, also sold phones to the Danish market. I bought the one I have from Denmark. If I had still been living in Norway, I am sure I would have been able to find a grey one. I saw one or two in antique shops there before I moved, and before I was interested in getting one. I hadn't quite reached the nostalgic age yet while I lived there. I've heard it's easier to use the speaker than the actual mic when converting it to a microphone. I didn't know that was because it was a condenser, though. I had always assumed it was dynamic, but that explains why it would be a bit more fiddly.

[@unotoli]

Don't they model it for VSTs? Should be well known fact / well tested. Roland got it on-board too, not sure about other digital pianos.

[@hakonsoreide]

Most piano VST's are sample based, not modelled. Sample-based is far easier to make than acoustic modelling, after all. Roland has a modelled piano sound in their electronic pianos and some workstations, but they are so far the odd one out in the hardware world. A rare few sample-based ones also include resonance samples. The Giant by NI is one of them, but the setting to make them sound isn't on by default, if I remember correctly, so you have to know about it to make the best of it.