����VR��Ƶ

Tone colour differences across diverse singing voices are more readily apparent than tone colour differences across players of the same instrument. Although experienced listeners can differentiate between performers of the same instrument, existing research suggests that there are greater changes in tone colour, or timbral variation, across different singing voices. Vibrato is commonly accepted to be a distinctive feature of singing, inextricably tied to expression and tone colour.[i] Previous studies have shown that the addition or subtraction of vibrato changes people’s perception of the timbre of the sound. [ii, iii] Therefore, vibrato constitutes a significant component of tone colour and may be a significant distinguishing factor of timbral variation in different singing voices.

The purpose of this research was to gather information about how singers and instrumentalists experience and describe the timbres—or sound qualities—of different singing voices. One study (from two overall) is reported concerning timbre qualia (description semantics) of the singing voice. Across the study, open-ended interviews were conducted with 30 participants, including 15 vocalists and 15 instrumentalists. The participants were asked to use as many adjectives or descriptive words (without prompting) to express their experiences of the timbres of unaccompanied vocal excerpts sampled from three genres (classical, jazz, musical theatre). These unaccompanied vocal sample data were gathered from a singing voice protocol with genre-fluid repertoire designed specifically to target style-specific vibrato from the individual singers.

Selective content analysis of the interviews suggests discrete qualitative categories underlying the participants’ descriptions. Further analysis examines hypothesized differences in descriptive strategies in the survey participant demographic between instrumentalists and vocalists regarding attention to vibrato variability. The results of the study are compared to those of Reymore and Huron (2020), who collected descriptions of musical instrument timbres. This research raises awareness of the audience’s perception of vocal timbres within the same material performed with stylistic variations, especially vibrato. This benefits all musicians in expanding their educational practices, considering the effect of vibrato variability on timbre and audience tone colour reception for optimal singing, compositional, and artistic practices.

• i Sundberg, J. (1987). The Science of the Singing Voice. Northern Illinois University Press, Dekalb, IL, 170.
• ii Almeida, A., Schubert, E., & Wolfe, J. (2021). Timbre Vibrato Perception and Description. Music Perception, 38(3), 282–292.
• iii Loni, D. Y., & Subbaraman, S. (2018). Timbre-Vibrato Model for Singer Identification. Information and Communication Technology for Intelligent Systems, 279–292.

Deep learning models have provided extremely successful methods in most application fields by enabling unprecedented accuracy in various tasks. For audio applications, although the massive complexity of generative models allows to handle complex temporal structures, it often precludes their real-time use on resource-constrained hardware platforms, particularly pervasive in this field. The lack of adequate lightweight models is an impediment to the development of stand-alone musical instruments based on deep models, entailing a significant limitation for real-life creation by musicians and composers.

The main goal of my PhD is to develop specifically-tailored approaches that could allow to enable a larger use and a simpler control of deep generative models. This lead to the development of completely novel tools that would allow to democratize deep models on innovative instruments, thus initiating a new path for creativity.

Hence, we recently built the first deep learning-based music instrument by implementing a lightweight generative musical audio model on an adequate hardware platform that can handle its complexity. By embedding this deep model, we provide a controllable and flexible creative hardware interface. More precisely, we focused our work on the Eurorack synthesizers format, which offers Control Voltage (CV) and gate mechanisms allowing to interact with other classical Eurorack modules. As a timbre case study, we decided to focus our research on impact sounds, as they are notoriously hard to create with traditional audio synthesis methods. In the future, we aim at developing similar deep-based hardware instrument, tackling the problem of mixing and orchestration for electro acoustic music.

Olivier Messiaen, famous for having sound-color synesthesia, describes his complex yet consistent colour system in his Traité de Rythme, de Couleur, et d'Ornithologie (1994). Although much existing research focuses on his personal colour system, one finds evidence in his descriptions that he also draws from crossmodal correspondences, or intuitive and unexpected connections between sensory modes shared by most people and even some animals. This isn’t surprising, as research suggests synesthetes experience crossmodal correspondences more strongly than do the general population, yet such a claim about Messiaen has not yet been made. If Messiaen drew from crossmodal correspondences in addition to his synesthesia, it could dramatically change the way we listen to his work. Elements of his style which had been deemed inaccessible to most listeners would not actually be so, as we can experience the sensory elements which come from crossmodal correspondences: the spectrum of darkness to light, of warm to cold, of saturation or intensity of colour, and of texture and shape. Whereas his synesthesia is based on pitch class/chroma, when he ventures into the shared world of crossmodal correspondences, one finds an increasingly crucial role for both timbre and orchestration. I examine both Messiaen’s own descriptions of his experience and many examples of crossmodal correspondences he intuitively used in his published analyses of Debussy, connecting both types to published psychological experiments. I will also show evidence that Messiaen’s personal colours seem to be strengthened when they coincide with crossmodal correspondences. 

In my presentation I would present part of my master's thesis, which I wrote in 2021-2022. In my thesis ""Timbral emergence in orchestration after 1950"" I examine the different ways of generating a genuinely new timbre, the emergent timbre, in orchestral music after 1950.

The part of my thesis that I would be presenting for ACTOR is about using gestalt theory (gestalt psychology) principles as a technique of emergent orchestration to create genuine new timbres and timbral structures. I take the music composed after 1950, because in the post-war period the sound quality and timbre become the central idea of the compositions. The timbre was understood and applied as the primacy of the composition, while the traditionally more important musical elements, harmony and rhythm, would serve to underline it.

 The Gestalt theoretical disciplines I would examine are phase transition, grouping tendency, zero-point shift and primacy effect based on the pieces ""Serendib for 22 instrumentalists"" by Tristan Murail and Ligeti’s Violin concerto. Furthermore, I explain the formation of ""Virtual Instruments"", for which these principles are mostly responsible, with a partial analysis of the orchestration of ""4 Adagi per flauto dolce e orchestra"" by Salvatore Sciarrino. In this regard, I use both the research in the field of music and the research in the visual field of gestalt theory. 

My piece ""Etude for ensemble"" for seven performers was written in the context of the 2021-2022 Composer-Performer Orchestration Research Ensemble [CORE] Seminar at ����VR��Ƶ, directed by Profs. Stephen McAdams and Guillaume Bourgogne. This two-semester seminar takes the form of a research-creation project where performers and composers work together on a specific timbre and orchestration-related topic.

The piece explores the compositional idea of what I call “compound figures”, where motive-like units are formed by the perceptual aggregation of multiple gestures distributed across different instruments. This idea draws from several compositional and theoretical concepts, including Brian Ferneyhough’s figure/gesture. Gérard Grisey’s instrumental synthesis and Marco Stroppa’s Musical Information Organisms [OIM], as well as Dennis Smalley spectromorphology and Albert Bregman and Stephen McAdams’ work on auditory streaming.

For this project, my goal was to use timbre in a way as complex and flexible as I would normally deal with pitches and durations. To achieve this, I tried to think of timbre as a multi-layered complex rather than a single dimension. For example, I tried to apply the way chord progressions are built from chords and chords from individual pitches to the timbral parameter. While this approach  greatly favoured my use of timbre as a form-bearing parameter and not only as surface feature, it also raised many issues linked with performability, notation and perceptual factors. 

I propose to discuss these issues, the way I dealt with them and the potential for further exploration."

Timbre plays a crucial role in sound source recognition. Musical instrument sounds carry timbral information about two mechanical properties: the excitation sets into vibration the resonator, which filters sound components. Excitation–resonator interactions in the physical world are restrictive: Strings can be bowed and struck but seldom blown. We used Modalys to combine three excitations (bowing, blowing, striking) with three resonators (string, air column, plate), simulating nine interactions. These interactions are typical (e.g., bowed string) or atypical (e.g., blown plate). 

Experiment 1 involved dissimilarity ratings of stimulus pairs. Experiment 2 entailed explicit categorization of excitations and resonators of the stimuli. Experiment 3 comprised three learning tasks training participants on the stimuli’s excitation, resonator, or interaction categories. Training involved trial and error with corrective feedback, followed by an explicit categorization task based on trained categories.

Multidimensional scaling revealed a three-dimensional timbre space (Experiment 1). Dimension 1 showed a clear boundary between struck and continuous excitations. Dimension 2 isolated plates and Dimension 3 further separated strings and air columns. Listeners accurately categorized excitations and resonators of typical interactions (Experiments 2 & 3). They assimilated atypical interactions to typical ones (Experiment 2). This confusion was reduced after training was involved (Experiment 3). 

Therefore, categorical boundaries of excitations and resonators were already formed implicitly and not made explicit until training took place. These studies reveal that excitation and resonator properties can be processed independently. Furthermore, they highlight the role of timbre in an essential process of human behaviour: identifying a sound source.

Creating timbre through blending of individual instruments is an effective and affective orchestration technique which critically shapes music perception. However, it remains mostly unclear how listeners perceive musical instruments in complex sound mixtures presented under realistic acoustical conditions such as concert hall acoustics. Using Wagner’s prelude to the opera Tristan and Isolde as a case study in my PhD research project, instrument identification performance serves as a proxy for quantifying perceptual effects of musical and acoustical attributes in musical scene analysis.

Individual (mostly) dry instrument tracks taken from OrchPlay will be used for rendering the complete sound mixture with realistic concert hall acoustics. The toolbox for acoustic scene creation and rendering (TASCAR) will allow to build a reverberant acoustic scenario of a concert hall. A loudspeaker array installed in an anechoic chamber will provide surround-sound playback, effectively creating the virtual acoustical presentation, placing the listener within the scene. Using short excerpts from the score with varying musical (e.g. number of instruments, stratification) and acoustical (e.g. distance between instruments on stage, reverberation time) parameters, the instrument identification performance of human listeners will be quantified in psychoacoustic experiments, where individual target melodies have to be attributed to particular instruments within the mixture. Furthermore, the sound quality and blend properties of different acoustical configurations will be assessed.

First results from the virtual acoustic scene will provide a benchmark for future experiments and modeling approaches on musical instrument identification to uncover the acoustical attributes that shape musical scene analysis in realistic multi-source scenarios.

Timbral blend is fundamental in various musical activities for shaping sounds and musical intentions. Previous studies on blend perception have mostly focused on sounding blend, neglecting the presence of imagined blend made possible by inner hearing. To investigate how imagined blend compares to the perception of heard blend and if musical background makes a difference in it, two groups of participants (musicians and non-musicians; 31 per group) were presented with pairs of short instrumental sounds in unison from 14 different instruments in two different experimental conditions. In the first condition, paired instruments were played sequentially, and participants were instructed to imagine them being played simultaneously and rate their degree of blend. In the next condition, pairs of instruments were played simultaneously, and participants were asked to rate the perceived degree of blend. Results showed significant interaction effects among the type of instrument pairs, presentation conditions, and musical backgrounds. Acoustic modeling and multidimensional scaling of blend ratings showed both varying and invariant roles of different acoustic features between the two types of blend perception. It appears that imagining blend is more sensitive to differences in brightness and richness of high frequency contents between paired sounds. A follow-up experiment on the perception of dissimilarity between instruments using the same stimuli provided further evidence that evaluating imagined blends is strongly informed by judging the dissimilarity of blending instruments. In practice, how the two types of blends differ is a result of complex interactions involving the specificity of blending instruments and listeners’ musical backgrounds.

In Luvale (and related) communities of northwestern Zambia, ancestors live on as makishi: elaborate spirit masquerade. The dozens of makishi are differentiated by their appearance, personality, and action. Many makishi speak, sing, or make noise and proper vocal timbre is vital to this performance (Euba 1988). In this presentation, I analyze the timbre of makishi voices by investigating the ways in which they are described. Interviews conducted during extensive on-site field research demonstrate that there is localized, insider vocabulary used to describe these timbres (Fales 2018; Feld et al. 2004; Wallmark 2018). I organize these words and phrases into Wallmark’s (2019) seven groupings of timbre descriptors, revealing that timbre is often discussed in similar ways across cultures. Both my methodology and focus on descriptors inherently present a listener-centered approach to vocal timbre that extends beyond the vocalizer (Eisheim 2018, 2019). This not only expands the topic of timbre to include perception, but also ensures that audience reception is considered a component of masquerade performance. Although performer agency allows for variation, makishi embodying the same ancestor are still required to adhere to a specific timbral archetype. Following Samples (2018), I argue that the presence of these group-specific descriptors demonstrates the archetype’s perceived distinctiveness. And the listener’s ability to map these timbral quale onto identity (Meizel 2013; Neal 2018) evinces cultural knowledge.

Lever du jour from Tableau III of Maurice Ravel’s ballet Daphnis et Chloé (1912) is an example of how orchestration can be used as a primary parameter to convey the image of daybreak (Millard 2021). Ravel uses techniques to transform orchestration such as alteration, addition, and expansion to gradually depict the climactic moment of the sun appearing from behind the horizon, accompanied by singing birds (Soden 2020, 71).

Daphnis et Chloé was originally choreographed by Michel Fokine for the 1912 premiere in Paris. Several choreographers have reimagined the ballet, including Sir Frederick Ashton (1951, London), Jean-Christophe Maillot (2010, Monaco), and Benjamin Millepied (2014, Paris). That Ravel’s score and Longus’ original novel of the same name continue to be of interest to choreographers speaks to the compelling partnership between the score and the extramusical program (Goddard 1926). 

In this presentation, I analyze Ravel’s Lever du jour to describe and quantify how ballet choreography may emphasize or contradict the trajectory of the scene as established through orchestration. I use three analytical approaches: 1) score-based analysis describes pitch and register content, and the introduction and elimination of instruments, 2) audio analysis accounts for sound intensity and timbral descriptors (such as spectral centroid and spectral flux), and 3) choreographic analysis (performance videos) adapts existing choreomusical notation systems to transcribe and analyze choreographies by Ashton, Maillot, and Millepied to study orchestration (Leaman 2016). This presentation aims to begin a conversation about applying quantitative analytical methods to choreography as a visual representation of orchestration.

Emotional impacts are one of the greatest motivating factors for people to engage in music. Previous studies have highlighted the significance of individual instrument timbre in conveying emotions. However, as a result of orchestration, individual sounds can be combined into a perceptual blend with new timbral features. How these features relate to perceived affect requires further study. An interdisciplinary research approach—a music perception study involving music, psychology, and acoustics—was adopted. Two experiments were conducted. The stimuli were 10 individual instrumental sounds and 45 instrumental blended dyads created from them. In the first experiment, 42 participants provided affect ratings on valence, tension arousal, and energy arousal for both individual and blended sounds. The second experiment, involving 40 participants, assessed how well the 45 dyad sounds blended. Statistical tests, acoustic analyses, and a novel geometric analysis were employed for data interpretation. We found that blends with different combinations of instruments are useful to express diverse affects, where blends with higher spectral centroid and spectral spread tend to have higher arousal. Notably, the affect of blends can be different from those of their constituent sounds. We also found that when two instruments blend, they have different contributions to the perceived emotion, and good blenders are always dominated by other instruments with higher spectral centroid. Additionally, we demonstrate that people with different levels of musical sophistications perceive emotions in a similar manner. In conclusion, this study attests to the importance of timbral and orchestration features in effectively conveying emotions.

A hybrid genre produced in colonial Shanghai beginning in the 1920s, shidaiqu combined Chinese operatic and folk traditions, jazz, and Tin Pan Alley song. Musical elements of shidaiqu reflected pre-WWII Shanghai cosmopolitanism, as shown by scholars Andrew Jones (2001) and Szu-Wei Chen (2007). Chinese leftists labeled shidaiqu “yellow music,” a pornographic and decadent sonic signifier of Western influence, subsequently banned in 1949 by the Communist Party. Here, timbre remains unexplored as a locus of aesthetic and political change and embodied expression of hybridization. To explicate the relationship between the Chinese political landscape and popular culture, I analyze the vocal timbres of shidaiqu performers.

Using spectrographic case studies, I show how Chinese singers moved away from markedly “Chinese” inflections toward a timbral cosmopolitanism characterized by open-throated, breathy phonation. In Zhou Xuan’s 1936 “Express Train,” her vocal timbre exemplifies the “untrained” voice characterized by nasality, portamento motion between notes, and irregular vibrato. Xuan’s timbre in “Shanghai Nights” (1946) indicates an orientation towards a cosmopolitan audience, her previously bright timbre softened into a darker bel canto style. I extend this analysis to other famous shidaiqu performers, including Yao Li, Bai Hong, Gong Qiuxia, and Wu Yingyin. I use spectrograms to visualize the trajectory of shidaiqu vocal timbre over the course of their careers, spanning the 1930s and ‘40s, and identify key moments that stimulated these performers to make strategic alterations. Lastly, I discuss the reception of shidaiqu vocal timbre as markedly “yellow,” illustrating the intersections between the voice, culture, and politics.

In this talk, I'm presenting a framework for computer-assisted orchestration that combines music theory and computer science to facilitate the collaboration between artists and computational algorithms to orchestrate existing musical pieces. The process has been developed and tested for classical and romantic style orchestration. This structured approach unfolds across three pivotal steps, accommodating both human and algorithmic contributions in distinct roles.

The creative process starts with an in-depth analysis of the composition to orchestrate. Particularly pertinent in this phase is the identification of the different voices and of the section boundaries between the evolving textures in the piece. The result of this phase is an abstract score, called layer score, which contains all the separated voices (layers) present in the original piece, analyzed for their role in the composition. The orchestrator is also free to add other layers to this score if doing so is careful in respecting the style of the author.

The second phase of the process consists in the creation of an orchestration plan, outlining the instrumentation of each layer across distinct sections of the piece. The artist can iteratively review and select the plan outputted by a highly customizable probabilistic Markov model.

Finally, the orchestral score is constructed, by writing the notes contained in the layer score for the instruments indicated by the plan, taking care of the instruments ranges and playing techniques. This phase can be performed by hand or with the help of a Large Language Model, currently in development.

Many instruments used in traditional Nigerian Igbo music do not produce specific pitches. As such, their functionality in orchestral contexts is often understood as rhythmic. With such an understanding, the organisation of instrumental timbres that characterises this music risks being overlooked, or misinterpreted as primarily rhythmic in nature. While rhythm is, no doubt, critical to Igbo music, timbre is the primary focus of the instrumentalists who perform it. In this presentation, I argue that the basic function of the instrumental parts in mixed Igbo music (i.e., instruments and vocals) is not rhythmic, but timbral, in their extension of this music’s vocal lines. I present transcriptions and analyses of two pieces of Akwu-eche-enyi Igbo music, complemented by interviews with its practitioners based in Anambra State, Nigeria. This material enables exploration into how the pitch collection of vocal lines structures instrumental orchestration in Igbo music. Specifically, the collection and ordering of vocal pitches determine the organisation of timbres and phrasing within the instrumental part. Hence, in a mixed ensemble performing Akwu-eche-enyi, the instrumental part functions neither as accompaniment nor as a collection of rhythmic motifs. Instead, the rhythmic surface arises from the timbral relationships between instruments, and how they blend with voices to create a holistic musical work. By tracing the interconnectedness between voices and instruments in Akwu-eche-enyi, I shift conventional scholarly focus on Igbo music from rhythm to timbre, more effectively connecting the creative processes of its musicians with listeners’ perceptions of orchestration.

This presentation gives a brief insight into my master's thesis research-creation project. I examine the unique problem space of pipe-organ augmentation, focusing on the issue of spectral analysis and synthesis. As part of this project, I have developed a synthesis server in python, called OrganLab, which serves to emulate and mutate pipe-organ based sounds, which is then be placed in dialogue with the acoustic instrument. The pipe organ is often considered the original synthesizer, and I extend this rich tradition with additive, subtractive, and FM synthesis, allowing me to access effects not possible with the original instrument, like interpolation between stops, glissandi, and inharmonic sounds. Since each partial is independently controlled, the harmonic spectrum can be exploded, contracted, and distorted, creating a rich palette of new timbral possibilities. These innovations are put into practice with the piece Élégies, written for my hyper-organ interface at l’église Saint-Édouard, where I've been organist since July of 2022. Based on the 10 Duino Élégies of Maria-Rainer Rilke, the piece incorporates the aural iconography of the space, making use of the sounds of bells, the fire alarm of the church, and the sounds of footsteps through its many corridors. This symbolic and spatial exploration mimics the aural exploration that seeks to navigate the continuum of acoustic and simulated--the sacred and the profane.

The objective of my research was to develop conceptual and empirical tools for defining auditory equivalents of visually perceived form and material. To establish an empirical basis for my hypotheses, I anchored my research in the well-studied ‘similarity with enlargement’ model (Goldmeier 1936)—a framework broadly recognized for its ability to empirically distinguish between visual form and material. I developed an auditory analogy to this model, tested it empirically, and found that the results parallel those observed in vision.
Investigating the auditory equivalents of visual form and material presents significant challenges due to differences between the modalities. For example, in vision, form and material are associated with structures in two- or three-dimensional space, while in audition, time and perceptual aspects of sound—such as pitch, loudness, and duration—are one-dimensional. For these and other reasons, a direct analogy between auditory and visual form and material is not feasible. To address this, I defined visual form and material in terms of perceptual features that can be explained irrespectively of visual space. By examining how these features manifest in the auditory domain, I was able to define the auditory equivalents of visual form and material.
The research places particular emphasis on discussing the notions of timbre and texture. Although these phenomena may seem disparate—timbre is associated with one unified sound while texture can be associated with the collective sound of an ensemble—they share distinct perceptual features, making them key examples of auditory material.