Clone the project

git clone https://github.com/bdvllrs/ear-trainer.git && cd ear-trainer

For the icons, download fontawesome and put in the project directory.

The excerpt are generated by sampling from a collection of scores. To generate the scores, you will need some midi files.

Add some midi files in the midi folder. (You can for example get this dataset: https://www.kaggle.com/soumikrakshit/classical-music-midi)

Start python script

python make_file_list.py

This will generate a scores.json file containing all midi excerpt.

Once this file is generated, the midi files are no longer required for the app to work.

The sound is generated using a soundfont from MIDI.js. (https://github.com/mudcube/MIDI.js/blob/88d85c14165356d3fe26f84b30f2061676433da4/examples/soundfont/acoustic_grand_piano-ogg.js).

You can change the soundfont by updating the soundfont.js file.

• Tone.js for sound generation.
• Vexflow for scores.
• MIDI.js for the soundfont.
• PitchDetech live audio recording code heavily inspired by this repository.
• Pitchfinder to detect played pitch. Using the Yin et al. algorithm.