Strategies for building AI-enhanced audio software with impact
Tools
Tools
Yee-King, Matthew and d'Inverno, Mark. 2024. 'Strategies for building AI-enhanced audio software with impact'. In: AIMC 2024. Oxford, United Kingdom 9 - 11 September 2024. [Conference or Workshop Item]
|
Text
COM-d'Inverno2024a.pdf - Published Version Available under License Creative Commons Attribution. Download (310kB) | Preview |
Abstract or Description
In this paper, we explore the challenge of increasing the uptake of AI-music technology research within academia and across the music industry. We consider two key audiences for AI-music technology: other researchers (an effect known in the UK as ‘academic significance’) and amateur and professional music practitioners (known as ‘impact’). We review previous work investigating the interactions between code repository design, repository usage statistics and citation count. We look at previous work exploring reasons for the low uptake of AI-based and other music technology research, such as poor interoperability and lack of alignment with the needs of creative practitioners. We then present a preliminary analysis of 93 AI-music-related GitHub repositories wherein we examine the interaction between repository features and uptake metrics such as citations and downloads. We find that AI-music research code repositories providing downloadable releases of plugins, which inter-operate with existing music technology, can achieve high download rates. We also verify the previous finding for non-music AI repositories that the number of forks positively correlates with the number of citations for associated research papers, noting that the number of forks is related to the design of the repository. We end the paper by describing how we are currently developing our AI-music software using a combination of C++, JUCE, PyTorch, RTNeural and plugin technology, a setup we have chosen with an aim to increase significance and impact. We connect our development setup to the findings from the review of existing work and our GitHub analysis.
|
Item Type: |
Conference or Workshop Item (Paper) |
||||
| Data Access Statement: |
The work described in this paper used data that is available via the internet, such as Google Scholar pages and research papers. It also used data accessible via the GitHub API, namely statistics relating to repository activity, forks and so forth. |
||||
| Keywords: |
AI-music, JUCE, plugins, music technology |
||||
| Related URLs: |
|
||||
| Departments, Centres and Research Units: |
|||||
| Dates: |
|
||||
| Event Location: |
Oxford, United Kingdom |
||||
| Date range: |
9 - 11 September 2024 |
||||
| Item ID: |
37555 |
||||
| Date Deposited: |
19 Sep 2024 11:24 |
||||
| Last Modified: |
19 Sep 2024 11:27 |
||||
|
URI: |
View statistics for this item...
 |
Edit Record (login required) |