First samples are here – http://www.pianoid.com/sample.html
We hope it will be finally able to end use of piano samplers at least in classical musical education.
What do you think?

Just in my opinion, this one just as well as all the others while allowing a superior response to the player's actions as compared to the sampled pianos, still have an unpleasant synthetic aftertaste. I just can't shake this feeling off when listening to e.g. Pianoteq. It's good for many things, many people would prefer it to a sampled piano but I can't really say it is superior in every aspect.

Sorry to hear you, it seems to me that the sound indistinguishable from the live piano.

The overall sound sounds rather restricted in frequency range. I know that the note range is limited, but the frequency range also sounds very restricted, both in low and high frequencies. However, it does sound promising. The attacks seem to have something that Pianoteq is missing.

Greg.

Would be better able to judge it without all the reverb that's been added.

Valery --

Is it possible that you are biased?

. . . Have you been involved in the development or marketing of Pianoid?

From _which_ "live piano" is the sound indistinguishable --

. . . a Steinway 6' grand ?
. . . a Bechstein 6' grand ?
. . . a Yamaha 7' grand ?
. . . a Bluthner ?

Each of these pianos has its own sound:

. . . Which one does Pianoid match ?

Charles Cohen You're right. I am involved in the development and marketing of Pianoid – Pianoid is a startup project.
Pianoid is a physical model of a GP. Our approach is that we use precise model with huge calculating power. We use difference scheme to solve differential equations of strings, and we use modal approach to sounding board. As a result we get correct behavior of our model. The parameters we upload to Pianoid are: strings tension, stiffness, mass and length, bridge position, sounding board modes, their shapes, frequencies and Q factors etc. These data obtained through measurements of a real GP. The resulting sound quality is dependent on measurement accuracy.
Give us a time and we will do model of
a Steinway 6' grand
a Bechstein 6' grand
a Yamaha 7' grand
a Bluthner
I hope so!

The best way I can find to describe this is "It sounds like you're hearing someone play the piano through an old telephone".

Nice, there is something in sound that promising, so I am very impatient for your pianoid. How long to wait for version 1 ?

Funny, that is exactly what I think whenever I am forced to listen to music on YouTube, because people don't care enough to upload lossless audio in place of YouTube.

valerykx, interesting development - thank you for posting.

May I ask what platform this software will eventually run on, and what kind of spec would be required for full 88-key modelling?

Kind regards,
James
x

Interesting you aren't just using something like an rpiZero or rpi2 for this project. If you want cheap and programmable, that's where it's at...and today they released a fix that dramatically improves onboard audio playback.

PMFJI --

Pianoteq runs OK on an Intel i5 chip, but is marginal on anything slower.

The Raspberry Pi doesn't have nearly enough horsepower to do this kind of real-time computing.

Is this product the same that appeared a some years ago?

Hi Valery,

Thank you for sharing the samples.

I think that evaluation would be more productive if you could find a way to record the sound with higher quality. Honestly the samples sound like they were recorded on a C-60 cassette, hiss noise included (I'm sure many of you youngsters won't even understand what that means :D)!

Did you perhaps record the sound as diffused by loudspeakers? In that case, by all means find a way to record the direct digital output of your FPGA model.

Regarding the sound itself, the only thing I will say from what I could hear is that you might want to check whether your sustain is not shortened too much from excessive damping. That's most likely only a model parameters question though, so it should be easy to get a longer sustain. Ah, let me add that I also had the impression that when exercising dynamics there was a kind of 'two level quantization' of the spectral signature, which is exactly what I wouldn't expect from a modeled piano.

I see where you are going, and encourage you to go ahead, but there's a lot of work to be done to compete HW digitals and, most importantly, with Pianoteq.

Keep up the good work!
Paolo

Kawai James Pianoid will be a non-expensive chip (about $30) with few SPI inputs for uploading coefficients and inputting MIDI events, and output for audio. This chip will not be programmable, it could solve only one task – calculate GP behavior. So it does not need any software platform. In final implementation, this chip can be integrated in existing devises, with their own operation system, or as standalone devise with FLASH (for keeping presets), WIFI (for downloading presets and making custom presets through website), MIDI input and audio output. In this case devise can use cheap microcontroller with small and light operation system like linux.

At a present time we use FPGA Xilinx xc7vx690t. It is very powerful programmable chip for ASIC prototyping. The full 88-keys version does not fit in this chip, so we need bigger FPGA. They do exists, but we need to find investors to make prototype and produce ASIC. According to our investigations, there is no problem to produce ASIC with necessary

Valerykx,

Why have your posted a wav file that is digitally over-driven? There are many times during the performance where the clipping distortion occurs. See Greza-3.wav (Sweet Dreams)

Apart from that the recording is muffled and dare I say a little like a "revived" recording from an early 78rpm record. It is not a good advert for the technology.

Ian

This is a really cool project. Hope it goes well for you.
Myself, I work with ASIC development

How many are you working on this?

Orco thank you very mutch. We are working 10 years

Marcos Daniel, yes the same