Sound Analysis-Synthesis

Sample Orchestrator

The Sample Orchestrator project aimed to produce new applications based on content management and manipulation of sample (isolated pre-recorded sounds) banks. Its production corresponds to the conjunction of two elements: the widespread availability of large sample banks on different supports (e.g. CD and DVD, online databases) but that present a limited number of applications (e.g. synthesizer samples); and recent scientific and technological developments in audio indexing methods and database management systems, making it possible to imagine new musical functions, calling on global methods of content-based management and manipulation of the entire corpus of available samples:

• Content-based management of the sample bank: automatic classification, search by similarity, concatenative synthesis based on high-level descriptions;
• Content-based audio processing (based on a previous analysis of characteristics of the processed signals): transposition, time-stretching, hybridization, morphing, etc. These processing functions can be divided in pre-processing functions that address the preparation of samples in the studio, and post-processing functions that intervene in real-time when the sound banks are used via a new generation instrumental sampler;
• Orchestration tool, integrated in the OpenMusic environment, which finds, based on models of musical knowledge (generally empirical for composers), the best combinations of sounds with a specific musical effect in mind (fusion/separation, re-orchestration from existing sounds, etc.).

Research activities on identified scientific and technological issues were planned as a part of the project: description, indexing, automatic classification of musical and sound contents, content-based sound processing, musical knowledge engineering and its application for orchestration, software environments for real-time audio processing.

This project was the subject of numerous scientific publications (2 doctoral theses, 10 articles in journals and books, 18 conferences), of an industrial valorization (the technologies developed in the project led to the creation of the MachFive 3 software program by Univers Sons, marketed by the American company Mark of the Unicorn), and several musical creations at IRCAM. The project was selected for the label "major scientific advance" by the ANR in its 2006 report.

Project Details

Participants