Dr Alec Wright

What problem are you trying to solve?

Large‑scale audio machine learning can unlock new and exciting musician‑centric tools for creative music and audio software. Text‑to‑sound generation techniques, such as semantic audio understanding, timbre transfer and intelligent effects control, can be used to create innovative creative tools. 

Yet most music software companies are formed from small teams without the specialised research expertise, data pipelines, evaluation practices or computing power required to build and maintain these systems. This capability gap slows innovation and keeps many potential ML/AI features inaccessible to these companies. The problem I’m addressing is this mismatch between the industry’s needs and the technical demands of modern audio ML.

What is your idea?

The opportunity is to develop sector-specific foundation models, which would be too resource-intensive (both in terms of compute and technical expertise) for small companies to develop. These can be licensed to small audio‑technology companies, enabling them to deliver innovative products without having to build ML infrastructure from scratch.