Table of numbers → sound events, one column per parameter.
Bright pixels → grains scrolling over time.
Each numeric column can drive a different sound parameter.
Each instrument has its own envelope. The export MIDI writes the program General MIDI corresponding, so the DAW opens the right sound.
Everything here is opt-in and off by default: the pitch stays mapped 1:1 to your data (maximum fidelity). These controls only touch the timing, the micro-variation and add an accompaniment voice derived from the melody itself. Every choice goes into the JSON, so the render is reproducible.
Small imperfections make the result less "robotic". They do not change the notes, only their micro-timing and dynamics.
Adds chords and/or bass diatonici, chosen by analyzing the most present notes in your melody in each time window. It's music derived from the data, non inventata.
The export reproduces esattamente what you hear (same graph, offline render). The events JSON is for traceability and reproducibility.
There is no "the very best" mapping: there is the one consistent with the question you ask.
Need something immediate? Use SONA Lite. Want the visual side too? See Data Storytelling.