From Tacit to Tokenised: Critical Perspectives on AI Text-to-Audio Interfaces

Charalampos Saitis (Queen Mary University of London)*; Fabio Morreale (Sony AI); Landon Morrison (University of Rochester); Ashley Noel-Hirst (Queen Mary University of London); Rebecca Fiebrink (University of the Arts London); Joseph Meyer (University of the Arts London); Courtney N. Reed (Loughborough University London)

• The registration for this workshop is managed by the workshop organisers. Please check the workshop's website page for more information
• Type: Hybrid
• Room Location: 206
• Morning (9am-1pm)
• Signup Deadline: June 9th AoE
• Contact Email: here

Abstract:

In AI text-to-audio interfaces, forms of tacit musical knowledge traditionally developed through embodied interaction are translated into tokenised linguistic descriptors that condition generative models. These models assume that language can serve as a proxy for structuring latent representations and producing musically meaningful outputs. However, musicians and researchers have long argued that music resists containment through verbal description, that it does not operate like a language and exceeds the limits of pure structure. This half-day workshop examines the functionalising of language in semantic cues for AI audio engines, which is moreover largely curated by a small number of multi-national industry giants, and its implications for digital musical instrument design and musical practice more broadly. Participants will engage critically with popular AI text-to-audio models alongside movement-to-audio interfaces to explore how alternative semantic mediation strategies for generating sound shape musical meaning and creativity. A collective open discussion will critically reflect on their cultural, epistemological, musical, aesthetic, and political-economic implications for the NIME and off-NIME communities.

Workshop website: https://comma.eecs.qmul.ac.uk/critical-text-audio-interfaces/