SIGNAS

The Special Interest Group on Neural Audio Synthesis @ QMUL's Centre for Digital Music

View My GitHub Profile

NASH: the Neural Audio Synthesis Hackathon

SIGNAS are proud to announce that on 18th-19th December 2021 we will host our first Neural Audio Synthesis Hackathon (NASH) as a satellite event to DMRN+16. The event aims to encourage cross-disciplinary collaboration in neural audio synthesis by encouraging the development of new techniques, tools, and interfaces, with a particular focus on creative musical applications.

Projects are particularly encouraged under the following topics:

This is not an exhaustive list, and all submissions under the umbrella of neural audio synthesis are welcome.

Click here to sign up for NASH 2021

Why a hackathon?

Neural audio synthesis is continually growing in popularity, with impressive new models released with increasing frequency. Whilst several artists (for example, Dadabots, Holly Herndon, and Hexorcismos) at the intersection of music and AI research have begun to apply this technology in their work, there is still much exploration to be done into the creative possibilities of neural audio synthesis. The technical barrier to entry remains high for artists and musicians; the capabilities of state-of-the-art models have been explored mostly on a narrow set of benchmark tasks; and the sorts of innovation seen in the control of image generation (e.g. CLIP-guided generative models) are yet to be paralleled in the neural audio synthesis community.

A hackathon provides a valuable opportunity for researchers, artists, developers, and others to explore new ideas in a low-pressure environment. It is an ideal setting for forming new projects, exploring risky ideas, meeting collaborators, and knowledge sharing. A hackathon can also provide an entry-point to a field for people from a variety of backgrounds and experience levels. We hope that NASH can encourage community exploration of these open problems and foster exciting collaborations and innovations going forward.

Details

Inspired by the HAMR hackathon’s 2020 edition, NASH will take place remotely in 3 phases.

  1. Brainstorming and team sign-up
    Fri 10th December to Fri 17th December
    A link will be posted to a shared Trello board where interested individuals can post project ideas and form collaborations.
  2. Hacking
    00:00:00 UTC Sat 18th December to 23:59:59 UTC Sun 19th December
    The hacking period will run for 48 hours to enable participation from all geographic locations. We encourage participants to sleep, eat, and go outside, and not spend all this time hacking.
  3. Voting
    00:00:00 UTC to 23:59:59 UTC Mon 20th December
    A 24 hour voting period will allow participants and DMRN attendees to view and vote on the submitted projects.

After voting is concluded, we hope to present the results in a session at DMRN+16. Depending on the number of submissions received, we may play all submitted project videos, or just those of the highest voted projects. In any case, a link to all submitted projects will be provided.

Rules

In order to be included in voting, all teams must submit:

  1. A demonstration video of two minutes or less
  2. A public source code repository (or steps to recreate the technical portion of the hack in the case of creative applications)

Submitting an interactive demo is also encouraged where appropriate, although this is not a requirement for entry.

Voting

Whilst the spirit of NASH is not competitive (and there are no material prizes to be won), hackathon participants and DMRN attendees will be encouraged to vote on the submitted hacks to allow the sharing of work, and to recognise great contributions.

As with HAMR prizes consist only of bragging rights and will be awarded in the following categories:

  1. Best idea/research direction
  2. Best presentation/demo
  3. Best code/implementation

Click here to sign up for NASH 2021

This site will be updated with further information as the hackathon approaches, and all registered participants will be updated via email. In the meantime, if you have any comments or questions about NASH, please email b.j.hayes (at) qmul.ac.uk