We recognize a beloved musician even before we can name the composition.
A few chords, a particular pause, a barely perceptible shift in rhythm — and recognition already arises within us:
It's him.
"Music is the shorthand of emotion."
— Leo Tolstoy
And, as new research shows, this shorthand has an individual handwriting.
But what constitutes the invisible signature of a musician? Can we measure what we are accustomed to calling the character, style, or soul of a performer?
Researchers from the University of Cambridge decided to test this with the help of artificial intelligence.
Twenty pianists — twenty musical worlds
On 17 August 2026, a study was published in the journal Nature Machine Intelligence titled Machine Learning of Artistic Fingerprints in Jazz.
The first author of the work, Hugh Cheston, and his colleagues analyzed 84 hours of music — 1 629 performances by twenty famous jazz pianists.
Among them:
- Thelonious Monk;
- Bill Evans;
- Oscar Peterson;
- Keith Jarrett;
- Chick Corea;
- McCoy Tyner;
- Ahmad Jamal.
The audio recordings were converted into digital transcriptions, similar to a player piano roll. They showed which note was played, when it sounded, and how long it lasted.
Then the models learned to distinguish performers by four dimensions of their playing:
melody, harmony, rhythm, and dynamics.
The best model correctly identified the pianist in 94,4% of cases. Another model, more interpretable for analysis, which examined the four musical dimensions separately, achieved an accuracy of 91,3%.
The artificial intelligence did not see the musician's face and did not know the title of the piece. It recognized the person by stable musical decisions repeated across different recordings.
What gives the musician away?
In the model that separated musical parameters, the strongest indicator of the performer turned out to be harmony — characteristic chords and ways of connecting them. Next came rhythm and melody, while dynamics provided the model with less information.
However, the results depended on the way music was represented, so they cannot be turned into a universal law about the superiority of harmony over all other elements.
And yet, it is precisely in choice that a person is especially evident.
One pianist unexpectedly replaces a familiar chord. Another shifts the rhythmic accent. A third holds a note or leaves a pause where the listener expects continuation.
Each decision may seem insignificant. But together they form a stable pattern—a musical fingerprint of the performer.
Even performing the same jazz standard, musicians create different worlds.
Because the performer conveys not only the written notes. They bring to the music their own memory, bodily rhythm, experience, temperament, and a way of hearing silence.
Where the notes end
A musical fingerprint rarely consists of a single striking device. More often it is a multitude of nearly imperceptible decisions:
- when to enter;
- which note to hold;
- which chord to choose;
- on which beat to place the accent;
- where to alter the movement of a familiar melody;
- how much space to leave for silence.
Individually, these details can easily be mistaken for chance. But together they create a recognizable pattern that returns from performance to performance. A kind of musical DNA.
Notes can be repeated. Presence—cannot.
Artificial intelligence does not measure the soul
It is important not to turn the research result into a beautiful but inaccurate statement. The algorithm did not "see the soul" of the musician and did not reveal the whole mystery of creativity.
The models worked with digital transcriptions of piano playing. This format well reflects the placement and duration of notes, but does not preserve in full the timbre of the instrument, the acoustics of the room, the quality of touch, and the finest nuances of pedaling.
Moreover, the system studied a specific set of twenty well-known jazz pianists. High accuracy within this experiment does not mean that the algorithm can flawlessly recognize any musician in the world.
But the study discovered something important:
individual style is not only a subjective impression of the listener. There truly exist recurring structures in it that can be seen and studied.
The authors released the model code and created an interactive web guide to pianists' styles. It allows one to explore the characteristic musical features of each performer and compare them with one another.
Can a machine find our uniqueness?
Such technology could help determine the possible authorship of unknown recordings, study the influence of musicians on each other, preserve musical heritage, and teach young performers to hear differences between styles.
But perhaps the most beautiful meaning of the study lies elsewhere.
Artificial intelligence, often accused of erasing individuality, here helps to discover it.
It shows: even within a common musical language, a person remains unique.
One instrument. Twelve notes. A familiar composition. And each performer creates their own universe.
Sound as an autograph of presence
A musician signs a work not with their name. They sign it with a touch.
With the time between two notes. With the way of entering a melody and leaving it.
With how they let silence continue what was said.
Two people will never play the same melody identically, because sound passes through different lives, different hands, different hearts.
And every time a person truly sounds, music speaks their name.
