How it works
Timbria's engine reads an image and composes a piece from what it finds. Here is the mechanism, at the depth we can state exactly.
1. Reading the image
The image is measured: colour, brightness, contrast, edges, the density of detail, and how light moves across the frame. A portrait is read from top to bottom rather than left to right, so the picture's own composition carries the piece.
2. The style
Timbria suggests styles that suit the picture, best fit first, from a catalogue of several hundred. Each style is a set of musical rules — harmony, rhythm, instrumentation — that the engine applies to those measurements. You can accept a suggestion or choose any other style.
3. The piece
Those measurements shape the piece as it unfolds: melody, rhythm, dynamics, and where it changes key. The music is calculated by Timbria's own engine, not produced by a model trained on recordings.
4. Listening
A piece plays in the studio, on a museum visitor's own phone, or from a link shared to anyone. Every surface plays the same rendered audio.
Where AI is used, and where it is not
An AI model suggests styles, suggests the starting key, tempo and instruments, and — separately — writes each piece's title and a short description.
Every note is composed by Timbria's own rule-based engine, reading the image. The AI never writes or plays a note.
Each AI answer is kept, so the same image in the same style is not asked about twice.
Same image, same music
The same image, in the same style, gives the same piece, for a given version of the engine. A different photograph of the same painting — different light, a different crop — gives a different piece: Timbria reads the file, not the subject.
Not always pleasant, on purpose
Fidelity, not decoration. A stormy or dissonant image should give a stormy or dissonant piece; smoothing every piece into something pretty would be the failure, not the feature.
See it for yourself
The fastest way to understand the mechanism is to hear it work on an image you know.