Question 1

Is my photo uploaded anywhere?

Accepted Answer

No. The image is analysed in the browser and the music is synthesised on the device through the Web Audio API. The file never leaves the device, there is no account and nothing is stored.

Question 2

Is the music random?

Accepted Answer

No. A checksum of the pixels seeds the pseudo-random generator, so the same photo with the same settings always produces the same piece, note for note. The Reading switch changes that seed when a different take of the same photo is wanted.

Question 3

How does a picture become notes?

Accepted Answer

Global features set the character: hue chooses the key along the circle of fifths, saturation and contrast choose the mode, brightness sets tempo and register. The frame is weighted toward the golden-section intersections, so a photo composed on the thirds is read the way it was framed. The picture is then scanned column by column from left to right, and the vertical centre of luminance in each column gives the melodic height of the matching bar. Chords, voice leading, cadences and form come from music theory, not from the image.

Question 4

Is this artificial intelligence?

Accepted Answer

Not in the current sense of the word. There is no model, no training data and no weights. Every rule is written by hand and readable in the page source. The genre is algorithmic composition, which is older than neural networks: Musikalisches Wuerfelspiel of 1787 assembled a minuet from dice rolls by table.

Question 5

Who owns the result?

Accepted Answer

You do. Nothing is generated from anyone else's recordings. The grand piano voice uses the Salamander Grand Piano sample set by Alexander Holm, released under CC-BY 3.0; if a piece made with that voice is published, credit the sample set.

Question 6

What can I save?

Accepted Answer

Lossless WAV, or MP3 at 320, 192, 128, 96 or 64 kbps. The piece renders faster than real time and the slider positions are baked into the file.