NIME2022
Requirements for hands-on session.
We propose three different hands-on session, can be performed consecutively or simultaneously, depending on the amount of interested people. If you are interested in one of these three sessions, please check the material pre-requistites listed below. The repositories gathering all the needed code can be found here : (added the day of the workshop).
Workshop 1 : RAVE & Max/MSP
Main referees : Antoine Caillon
Learn to embed spectral VAEs and RAVE into Max/MSP or PureData to explore them, and use it in your future patches.
Requirements :
More informations and resources are available here
Workshop 2 : Embedding on hardware
Main referees : Jean-Baptiste Dupuy,, Nils Demerlé, David Genova
Learn to embed machine learning models on RaspeberryPi, and whether how to remotely control them with an external webpage (from your phone, computer, or whatever), or to perform automatic generation for exhibitions / installations.
Requirements :
- Linux / Max / Wnidows computer
- RaspberryPi 4 (previous versions not supported)
- SD card (16/32Go)
- ethernet cable
- external keyboard
Workshop 3 : Exploring & hijacking neural synthesis models.
Main referees : Axel Chemla–Romeu-Santos, Constance Douwes, Hugo Scurto
Explore fix different pre-trained and wrapped models (VAE, RAVE, GAN, Diffusion, DDSP) directly in Python notebooks. After exploring manually the models, we will also try different methods to hijack / alter / transfer the models to generate alternative content. No skill in Python needed!
Requirements:
- Linux / Max / Windows computer
- Google Drive account
Bonus : Workshop X
Main referees : Axel Chemla–Romeu-Santos, Constance Douwes, Sarah Nabi
This workshop can be thought as the black diamond trail of the hands-on session, addressed to skilled users that would like to run over the three workshops in a short time. First, you will find a tutorial to implement your own spectral variational auto-encoder here : (added the day of the workshop). After that, you can catch a running session or roam the workshop support pages to export your model for Max/Pure Data, Raspberry Pi, or start hacking your model (it should train quite fast on CoLab). Of course, you will not be alone : we’ll be here to support!