Spectral plays a picture as sound. It reads a grayscale image from left to right, and each row of the image becomes a sine tone at its own fixed pitch, high at the top and low at the bottom. Bright pixels are loud, dark ones are quiet, so the picture's shapes turn into chords, sweeps and textures.
This is a research prototype: it plays two built-in images, Portrait and Ridges, and does not yet load your own pictures.
Controls
- Source and playback: Image picks the picture. Drone keeps it sounding without MIDI, Loop wraps the end of the scan smoothly into the start, and Scan Duration sets how long one pass takes, from half a second to 30 seconds.
- Frequency: Low and High Frequency set the pitch range the rows cover, spaced on a Log or Linear scale. MIDI notes transpose the whole range, with middle C leaving it as set.
- Image: Black Point silences the dim background, Brightness Curve makes the sound sparser or denser, Row Smoothing softens sudden changes in brightness, and Spectral Tilt tames or lifts the highs.
- Amp and output: Attack and Release shape how the sound starts and stops, and Level sets the output.
The frequency picture in the panel shows how the 48 rows are spread across the pitch range. It is a map of the settings, not a live spectrum.
Getting started
Load the Portrait Loop preset and listen to the scan in drone mode, then try Ridges Loop for clear, separate lines of pitch. Turn Drone off and play MIDI notes to transpose the picture.
Spectral is monophonic. Thin bright lines come out as clear tones, while large bright areas turn into dense, organ-like chords.
The idea is inspired by the open-source SpectroTrace; Spectral is an independent implementation.