Prism

Experiment for imagined repetition of speech and music.

Requirements

Managed with pixi; most dependencies will be automatically installed when tasks are invoked. Exceptions:

• music stimulus generation (stimuli/make_stimuli.py) requires timidity and lilypond, which are not available through conda-forge. For interactive listening via music21's Stream.show("midi") syntax, I also needed fluidsynth.
• rsync cannot be managed via pixi on windows. It's helpful to have it when loading the stimuli onto the experiment runner computer (via pixi run rsyncstims). The task pixi run getstims is an alternative that uses scp instead.

Setup

music21 has a somewhat complicated setup process. Running python -m music21.configure is supposed to detect various accessory programs, but doesn't always succeed. Here's a partial ~/.music21rc file that was working for me during development:

<settings encoding="utf-8">
  <preference name="graphicsPath" value="xdg-open" />
  <preference name="lilypondBackend" value="ps" />
  <preference name="lilypondFormat" value="pdf" />
  <preference name="lilypondPath" value="lilypond" />
  <preference name="lilypondVersion" value="2.24.3"/>
  <preference name="midiPath" value="synth" />
  <preference name="pdfPath" value="xdg-open" />
  </settings>

The program synth used as the value for midiPath is a small shell script somewhere on your PATH (e.g., at ~/.local/bin/synth):

#!/usr/bin/bash
fluidsynth -q --no-shell $@