In this hands-on online workshop, you will explore MIRLC, an easy live coding environment built in SuperCollider to query sounds from Freesound.org, and will be introduced to how machine learning can be useful to improve the live coding experience. The workshop is designed to accommodate both beginners and experts in programming using a team-based learning approach.
The learning outcomes of the workshop are:
- Get a sense of the practice of live coding (music live performance using code) by manipulating online crowdsourced sounds and the automatic use of feature descriptors obtained from Freesound.org.
- Get familiar with the application of neural networks, in particular a multilayer perceptron used as a classifier, to improve the practice of live coding with crowdsourced sounds.
- Be exposed to the main steps to solve a problem using machine learning techniques: the creation of a dataset, training a model, testing the model, and performing with / evaluating the model in an iterative cycle.
- Understand how to combine different technologies in SuperCollider to build a prototype for live coding performance.
- Get insight on a participatory design approach to designing a prototype for live coding performance.
At the end of the workshop, you will be able to:
- Use SuperCollider and the MIRLC2 library to retrieve sounds from Freesound.org based on a live coding approach.
- Use a trained model using the FluCoMa library to retrieve sounds that are based on personal musical taste.
- Train your own model using the FluCoMa library to retrieve sounds that are based on your personal musical taste.
- Analyse how to define a virtual agent that can react to the live coder inputs using the FluCoMa library.
- Explore creative strategies to perform with a virtual agent using machine learning for live coding.
An introduction to MIRLC: an easy live coding language to query sounds from Freesound.org
- Introductions (all) via chat - 10'
- Introduction to the workshop (Sam) - 5'
- Introduction to the project / team / software (Anna) - 10'
- Tutorial Freesound quark (Sam) 10'
- Tutorial MIRLC 2.0 (Sam) 20'
- Comfort break - 10'
- Creative exploration of the MIRLC2.0 tool (breakout rooms) - 25'
- Plenary (all) - 15'
- Closing - 10'
An introduction to machine learning in live coding and task 1: Train an agent to assist meaningful querying to Freesound.org
- Recap (all) - 10'
- Project presentation 2nd part: task 1 (Anna) - 15'
- Tutorial FluidMLPClassifier (Anna) - 10'
- Tutorial MIRLCa performance (Anna) - 15'
- Comfort break - 10'
- Hands-on MIRLCa performance (breakout rooms) - 20'
- Tutorial MIRLCa training (Anna) - 15'
- Hands-on MIRLCa training (breakout rooms) - 15'
- Plenary and closing (all) - 10'
An introduction to machine learning in live coding and task 2: Explore an agent that reacts to the live coder inputs when querying to Freesound.org
- Recap (all) - 10'
- Guest speaker: Iris Saladino - 30'
- Project presentation 3rd part: task 2 (Anna) - 15'
- Tutorial FluidMLPRegressor (Anna) - 5'
- Comfort break - 10'
- Breakout rooms: Reflections on machine learning tasks and live coding (All) - 20'
- Planning for videos / performance with Phonos / Toplap Barcelona / L'Ull Cec (plenary / breakout rooms) - 20'
- Final feedback and closing (all) - 10'