Physis Plus represents the natural and powerful evolution of our patented and well-known Physis (Physical modelling) technology. How does it differ from the current one? In our laboratory we have developed a technological combination of our proprietary physical models and an artificial intelligence algorithm we call MORIS© (“Multi-Objective Random Iterative Search”, a Viscount Deep Learning complex procedure (*1)(*2)(*3)) that has made it possible to achieve a new level of sound richness. This algorithm is able to learn and capture new acoustic features of typical organ sounds and provide estimates of the parameters of a pipe physical model, making our new physical models incredibly realistic both in terms of harmonics and keyboard feeling.
When constructing an organ stop, the first things to determine are the physical features of each individual pipe in every detail; for example the diameter, width and height of the mouth, the shape of the resonator tube – if any – the stopper and chimney, etc. All these elements combine to determine the timbre as the features of each pipe.
In our case the work is done ‘backwards’; we take a set of homogeneous reeds (a rank then) and let the deep learning algorithm listen to them, analysing every smallest feature in detail, until determining, with a good approximation, how the sound object that determined that timbre is made. This work, done by the MORIS©, is therefore comparable to the work of the organ builder (or rather the pipe maker): its output is a sounding voice, but it is still “raw”.
However, to obtain voices of good aesthetic and artistic quality, it is required the intervention of the musician and sound designer who, in a manner, is comparable to the pipe tuner, who takes the sounds one by one and refines them so as to make them sound good by each self, and also well harmonised with the other voices.
MORIS© technological process
Starting from an Input, the MORIS© will seize those features that contribute to the creation of that peculiar sound, it elaborates them with a series of rules, and it produces a “raw” pipe organ sound through the synthesis. Through a series of iterations, randomly varying the synthesis parameters, MORIS© modifies the set of parameters until the best representation of the input sound is obtained.
Nowadays one of the most challenging tasks in physically-informed sound synthesis is the estimation of model parameters to produce a desired timbre. Automatic parameter estimation procedures have been developed in the past for some specific parameters or application scenarios but, up to now, no approach has been proved applicable to a wide variety of use cases. Now our physical modeling technology development has set a new standard of sound reliability and flexibility.
(*1) A Multi-Stage Algorithm for Acoustic Physical Model Parameters Estimation, IEEE/ACM TRANSACTIONS ON AUDIO, SPEECH, AND LANGUAGE PROCESSING, VOL. 27, NO. 8, AUGUST 2019, Leonardo Gabrielli, Stefano Tomassetti, Stefano Squartini , Carlo Zinato, and Stefano Guaiana
(*2) Timbre Equalization of Wind Instruments Physical Models Using a Random Iterative Search Algorithm, J. Audio Eng. Soc., vol. 68, no. 5, pp. 364–376, (2020 May), Leonardo Gabrielli, Stefano Tomassetti, Stefano Squartini
(*3) INTRODUCING DEEP MACHINE LEARNING FOR PARAMETER ESTIMATION IN PHYSICAL MODELLING, Proceedings of the 20th International Conference on Digital Audio Effects (DAFx-17), Edinburgh, UK, September 5–9, 2017, Leonardo Gabrielli, Stefano Tomassetti, Stefano Squartini , Carlo Zinato