0rchid*

  • Sound Workshop

Orchid*, Ircam's assistance to orchestration software, designed and developed within the RepMus (Music Representation) team, creates instrumental mixtures as a solution to an orchestration problem consisting in perceptually imitating an acoustic target. 
Orchid* is the result of 15 years of research on computational orchestration carried out within repmus and constitutes a unique research and development ensemble in the field of orchestration, with 3 PhD theses, dozens of articles, and several computer softwares that have been widely used for the composition of important works of the contemporary repertoire.
The fundamental concepts that led to Orchid* have been discussed within Ircam's "Orchestration" research group, bringing together numerous researchers (Gérard Assayag, Xavier Rodet, Geoffroy Peeters, ...) and composers (Yan Maresz, Fabien Lévy, Joshua Fineberg, Tristan Murail, Mauro Lanza, ...) since 2004. Yan Maresz is at the origin of the idea determining the main functioning of Orchid* and he was subsequently the main musical advisor throughout the development of the software. As he says: "The work of this research group has resulted in a first concrete tool for musical research in the field of orchestration [...]. It is a question, in a very schematic way, of considering the problem in an "explorable" way through calculation thanks to the progress made in the description of sound. ...] It will also allow us to address the problem of electronic sound orchestration and that of "instrumental synthesis «  by getting closer to a given acoustic model, [...] by a finer approach by considering instruments as generators of complex sounds that can be mixed to obtain an acoustic or symbolic target. » (Yan Maresz, Pour un traité d'orchestration au XXIe siècle, L'étincelle, November 2006).

In this spirit, Orchid* is a complete system for computer-assisted orchestration and the optimization of timbre mixtures. It provides a set of algorithms allowing the reconstruction of any sound target evolving over time by a combination of instruments or samples, according to a set of acoustic and perceptual criteria. It can help composers to obtain unprecedented timbre colors by providing a multitude of effective solutions that best recreate this sound target. Thanks to an extensive set of features, Orchid* can also reproduce timbral and orchestral evolutions by analyzing temporal dynamics according to various strategies. Its results provide multiple orchestral scores that can be intuitively organized to quickly realize orchestral and musical ideas. Orchid* can use several databases of instrumental sounds but can also be extended in an unlimited way, including with synthesized sounds, by simply importing its own sound bank.

A complete bibliography of Orchid*'s work can be found here.

Orchid* was declined in three successive software instances developed within the team, each of which put forward specific aspects while trying to continuously improve musical quality through the choice of heuristics and the quality of musical experimentation.

Orchidée, the first realization set the laws of the genre in terms of functionalities and methods: imitative orchestration, prediction of audio descriptors of mixtures, multi or mono objective optimization by genetic algorithm for the search of orchestral solutions, static orchestration (single target) or sequential orchestration (by segmentation / chaining), mixture of partials, constrained filtering, etc. Orchidée was available as a standalone or as a Max interface toolbox (Orchis) or OpenMusic interacting with the orchestration engine.

Orchids, the second realization, extended the Orchid field, moving from multi-objective optimization in a descriptor space to time series optimization on these descriptors, paving the way for continuous dynamic orchestration of complex and evolving spectral morphologies. Orchids offers a rich standalone interface.

Orchidea, the most recent version, is based like the others on genetic research for single or multi-objective optimization and descriptor prediction, with a standalone version or a Max client. Among other things, Orchidea further refines and optimizes the analysis and prediction of descriptors, the genetic search for solutions and the quality of sequential orchestration, offers widely tested powerful musical heuristics and streamlines the user interface.

Orchidée was written by Grégoire Carpentier who devoted his doctoral thesis to this environment in collaboration with Damien Tardieu's thesis  in the Analysis - Synthesis team, which sets the descriptors forecast mechanism.

Orchids was written by Philippe Esling, who devoted part of his PhD thesis to this environment, and Antoine Bouchereau who finalized it within the framework of an Ircam R&D contract.

Orchidea was written by Carmine Emanuele Cella within Ircam's Musical Representations team under R&D contracts co-financed by Ircam and the Haute Ecole de Musique de Genève (HEM), a collaboration set up by the late Eric Daubresse to address the third generation Orchid*.  Carmine Cella continues today to participate in the project at the University of Berkeley where he is Professor, as well as creative contributors such as Daniele Ghisi.

From an institutional point of view, the rights (copyright) are divided between Ircam (which brings all the previous knowledge of the Orchid* line) and, for Orchidea, the HEM. Orchidea and Orchids are no longer maintained.

It is planned that the further development of Orchid* will take place within the framework of a collaboration between Ircam and the University of Berkeley, so that the software, which arouses the interest and commitment of several institutional players (Ircam, HEM, Berkeley) and renowned composers, can continue to evolve under the best conditions. To do so, we plan to release Orchid*'s code in OpenSource so that all interested people can continue to collaborate and freely benefit from their contributions and creations.

En poursuivant votre navigation sur ce site, vous acceptez l'utilisation de cookies pour nous permettre de mesurer l'audience, et pour vous permettre de partager du contenu via les boutons de partage de réseaux sociaux. En savoir plus.