Takafumi Horiuchi

Arpeggicho

A system that turns arpeggios into chords in real time

1. Overview

The system records each note you play and layers the recordings so you can hear an arpeggio as a chord. Even with an instrument that plays one note at a time, you can hear your own notes together and explore their harmony in real time.

Demo of the prototype under development

2. Background and challenge

While practicing arpeggios on the saxophone, I found myself memorizing them as sequences of finger movements and fingering patterns. I struggled to connect the notes I played one by one with the sound of the chord they would form together.

This led me to design a tool that layers the notes I play so I can hear the relationship between arpeggios and chords.

3. Approach

Left: C and E have been played. Right: adding G completes the C major triad.

Experience 1: Build a chord as you play

The system keeps playing the notes you have already played. When you play the next note while listening to them, the held notes blend with your live sound, letting you hear the intervals and the harmony as you perform.

Experience 2: Listen back after playing

After you stop playing, your recorded notes continue sounding together as a chord. You can listen back to the harmony your arpeggio created without having to play at the same time.

Key idea: Use your own sound

Playback uses your own performance, recorded on the spot, rather than preset instrument sounds. Your tone and intonation are preserved, so both building a chord while playing and listening back afterward remain grounded in your own sound.

4. Design and implementation

Because the system works with sound in a physical space, the environment and the conditions for audio input and output are part of the design. At startup, it measures ambient sound for about one second and uses that level to calibrate the threshold for detecting played notes.

The speaker plays back recorded sound while the microphone picks up new notes in the same space. Since the microphone receives both, distinguishing playback from newly played notes also became an implementation challenge.

The system excludes segments with unstable pitch or volume. It selects a stable segment of 150–300 milliseconds from each recording and plays it as a smoothly joined loop. It also limits output peaks as notes are layered to prevent excessive volume.

5. Next steps

The system is still under development and has not been released publicly. I want to improve sound recognition and make the interface easier to use during practice. Its learning benefits have not yet been evaluated. Through actual practice, I plan to investigate whether it helps players hear the relationship between arpeggios and chords.