Search

Some 40 years after the birth of computer music technology, new developments have emerged from research institutions that have gradually come into public use. This article examines recent developments that have offered new possibilities in sound analysis and synthesis. While they are typically issued from laboratories, these possibilities have reached not only professionals, but also a large public that is making or simply listening to music. First, we examine additive synthesis, one of the oldest methods of making computer music. Additive synthesis is the summation of time-varying sinusoidal components. Its development and use have been discouraged by several drawbacks, including computation cost, difficulty of introducing precisely controlled noisy components, and the challenge of controlling hundreds of sinusoids. Methods have been designed however to automatically analyze sounds in terms of partials and noise that can then be applied directly to additive synthesis (Serra 1989; Rodet and Depalle 1992), and we have produced a new additive synthesis method named FFT-1 that is based on spectral envelopes and the inverse Fast Fourier Transform. We will explain how this method decreases the computation cost and renders the control more simple, direct, and user-friendly. Commercial applications should appear this year for professional users and soon after for the public (Oberheim 1994). After the additive technique, we will examine the new possibilities offered by processing natural sound and voice. The recreation of a castrato voice by the means of sound processing (Depalle, Garcia and Rodet 1994) has been done to produce the soundtrack for a film and a CD about Carlo Farinelli (or Carlo Broschi), the famous castrato of the 18th century. This example is particularly interesting for several reasons. First, it is extremely difficult to synthesize a high-quality singing voice for a concert or a CD (Bennett and Rodet 1989). Second, the producers were able to create 40 minutes of a new castrato-like singing voice in a limited amount of time. Third, this is the first application of a technique that is analogous to morphing. And finally, the CD and the film have reached millions of

The CHANT project was originally concerned with the analysis and synthesis of the singing voice. This work led to a complex program of voice synthesisby-rule: CHANT. This program was enriched with a constantly expanding software environment, consisting of both analysis and composition programs. In time, broader aims than the synthesis of the singing voice imposed themselves. These aims centered on the search for models of the processes involved in the production of musical sound. Our present research encompasses the physical description of sound phenomena (the sonic material), the articulation of these phenomena (organization), and compositional issues. This research is intended to transcend simulation. Our goal is to extrapolate new creative models for music on the basis of knowledge models developed using the synthesis-by-rule methodology. In this research, synthesis is the proof of both our understanding of sound phenomena, and of the music itself.

For some time, it has been recognized that the development and use of complex systems has been stifled by the inadequacy of ordinary programming languages (Winograd 1979). Music composition and synthesis (MCS) by computer offers a particularly clear example of this "complexity barrier." In the 1970s, a new generation of interactive languages appeared (Cointe 1983), establishing the concepts of active objects and message passing as a control structure for computation. It is significant that object-oriented programming matches most of the requirements of MCS. Toward a solution to problems of complexity in computer music, an object-oriented programming environment called FORMES has been developed and implemented at IRCAM. It supports a new language with original features for composition and scheduling of objects. FORMES "processes" are obviously not in exact correspondence with musical processes as perceived by the listener. FORMES is an interactive system first intended for MCS. However, by virtue of its unique architecture, its flexibility, and its ease of use, it should find applications in a large number of domains such as graphics animation and speech synthesis. Composers use FORMES for score development and sound synthesis. As an example, Chreode, a piece composed with FORMES by Jean-Baptiste Barriere, won the International Bourges Festival award in 1983 (Barriere 1983). Why is MCS one of the most interesting fields for testing advanced technology for computing? In MCS, perhaps even more than in artificial intelligence, programs have to be tested, modified, and rewritten very often. This modification must be effected quickly and easily; otherwise, the continuity between musical ideas and results gets lost. The complexity usually involved in human composing activity is far beyond what we are nowadays able to code in a few weeks or months. At the same time, scientists and musicians cannot be hampered by esoteric or "hairy" programs in the course of sophisticated musical research. Thus, "friendly" interaction is not a wish but a necessity. Finally, the system must be interactive and fast enough to control a real-time synthesizer.