quote:

---

Originally posted by Russ2k2
Description
Chant was originally designed for the analysis and synthesis of the singing voice and was then found well suited for instrument simulation and synthesis in general. This technology is available in the jMax, Diphone Studio and Max/MSP (IRCAM/Cycling'74) environments and can be easily ported to other hardware or software systems. Independant from Chant, the Resonance Models provide unbeaten sound quality and effects for percussive sounds


Application
Real-time sound synthesis
Synthesis with Chant is flexible enough to be used in real-time. For instance Chant has been extensively used in virtual reality installations produced by IRCAM such as "Le Messager" and "Alex" (Catherine Ikam, Jean-Baptiste Barrière). In this context, visitors were able, with their motion, to interact with an avatar processed in real-time and to control the parameters of the avatar voice generated with Chant. Chant has also been used at NCSA (University of Illinois) in order to produce and control 3D sound spectra in real-time
Sound design
the analysis process for Chant is fully accessible within IRCAM's Diphone Studio in order to create unusual sound effects starting from different kind of sounds.. Live interaction : for concerts, Chant patches are available in the jMax and Max/MSP environments.
Features
Chant Synthesis
In order to explain Chant, the best way is to get back to the sound production model of the voice. The vocal track is composed of an energy source and a set of resonators. The lungs produce a flow of air which is turned into a signal by periodic (vowels) or chaotic (consonants) modulation. This signal is then filtered by the system (larynx-pharynx-mouth) which gives a timbre to the phoneme.
Chant simulates this with an excitation-resonance model. To each resonance or formant is associated a basic response simulated with a specific synthesis technique, the Formant Wave Functions (FOF). Chant produces the resulting sound by adding the FOF corresponding to each formant for a given pseudo-periodic source. In parallel, filters can be used in the same way to filter a sound.
The Chant analysis allows to define the response of a certain number of phonemes and resonant instruments and to characterize their specific temporal variations. This relieves the composer from searching the proper parameters and provides him a simple and intuitive control.


Resonance models
First conceived as an application of Chant, it is an analysis/synthesis method developed for modeling impulsive sounds (percussion, buzzs, pizzs, etc.). This method extends previous studies on timbral interpolation, and brings a continuity between synthesis and processing. These models (which can be considered as filter banks where frequencies, amplitudes and bandwidths are controlled) can be used to control synthesis (FOFs or filters with a noise impulse), and transformations (filters with an external source). The ResAn plug-in in Diphone Studio is meant for precise analysis according to this model. Then, synthesis can be performed in Diphone Studio or in Max/MSP using CNMAT's Resonating filters.

Analysis and edition with Diphone Studio
Diphone Studio brings access to a set of parameter for the analysis and provides a breakpoint function editor allowing very accurate interpolation of analysis segments (called Diphones). Diphone is the graphical interface for editing and interpolating FOF segments in the Chant plug-in.



Participants
Design and development: Xavier Rodet, Yves Potard (Analysis/synthesis team)
jMax version : Norbert Schnell (Real-time systems team)
Resonance models : Pierre-François Baisnée, Yves Potard, Jean-Baptiste Barrière
Diphone Studio version : Xavier Rodet, Adrien Lefèvre, Dominique Virolle

Configuration
Chant and Resonance Models, as Diphone plug-ins, runs on MacOS9. In real-time, FOF objects and Chant patches are available in jMax (Irix, Linux, MacOSX and Windows) and Max/MSP (MacOS9).

---