Workshops
Workshop 1
Dialoguing with Improvising Machines: A comparative-phenomenological Workshop on Musical Agent Systems by Nicola Leonard Hein (University of Music Lübeck), Pierre Alexandre Tremblay (Conservatorio della Svizzera italiana), & Gilbert Nouno (Haute école de musique de Genève)
This workshop brings together composers, improvisers, performers, creative coders, and researchers to critically explore musical agent systems and their musical affordances. It addresses a central question in contemporary computer music: how musical agents encode particular modes of listening, interaction, and agency, and how these design choices shape performance and composition.
The workshop is led by Nicola Leonard Hein, Gilbert Nouno, and Pierre Alexandre Tremblay whose artistic and research practices span composition, improvisation, creative coding, and interactive systems.
Adopting a commented comparative-phenomenological approach, the workshop will examine four musical agent systems through presentation, performance, collective listening, and discussion. A particular emphasis will be placed on the diversity of systems developed through different programming languages, creative coding environments, and technical frameworks, and on how these technological choices condition aesthetic behavior and interactional possibilities.
Each selected system will be explored in a 30-minute format consisting of:
- a short presentation of its technical design, musical aims, and underlying assumptions;
- a performance by the creator of the system;
- two short exploratory performances by peer musicians;
- and a critical discussion with participants and attendees.
The concluding session will open a broader comparative discussion across all four systems, with the aim of developing shared vocabulary, critical perspectives, and possible evaluation criteria for improvising and interactive musical agents.
Workshop 2
Computer-Assisted Composition Tutorial: From Survey to Studio Workflows with MIDI-GPT by Philippe Pasquier (Simon Fraser University)
Computer-Assisted Composition (CAC) is a long-standing field concerned with the use of computational systems to support musical composition. A recent survey of 125 systems (1963–2025) highlights both the historical continuity and rapid evolution of CAC. Early systems relied on rule-based and stochastic processes, while contemporary approaches increasingly employ machine learning, particularly deep and transformer-based architectures, as well as diffusion models.
This work also emphasizes that CAC systems should be understood not only in terms of their generative capabilities but through their interaction paradigms, including one-shot generation, iterative refinement, and co-creative dialogue. CAC is therefore distinct from autonomous music generation in that it explicitly centers the role of the human composer.Despite this, current AI music tools, often live in the cloud, privilege prompt-based, one-shot interaction, and are trained on ethically questionable datasets. While accessible, these approaches can reduce compositional agency and limit iterative exploration. This raises concerns regarding control, expressivity, and interaction design.
This workshop responds by focusing on studio-integrated, iterative workflows through the presentation of MIDI-GPT in Reapper and Fragmenta, positioning AI as a compositional operator embedded within Digital Audio Workstations (DAWs), rather than a black-box generator.
Workshop outline (2h total):
- Introduction and motivation (10 minutes)
- Overview of the CAC typology (20 minutes): https://doi.org/10.1145/3812539
- Transformer-based models for symbolic music generation, GigaMIDI and MIDI-GPT (20min)
- Examples of systems and studies (10mins)
- Interactive Demonstration in Reaper (10min): will demo the various functionality of the system and go through common use cases.
- Hand-on generation and discussion (20mins): participants will be able to try it for themselves (using a web-based version in Calliope to cut installation time). Participants will listen to each other’s generated snippets and discuss affordances, excitations, frustrations, and other reactions.
- Demonstration of Fragmenta and discussion (20 minutes)
- Conclusion (10 minutes)