Thoughts on Generating Music with AI

Up to now, artificial intelligence that aims to produce ›music‹ means little more to me than something to toy around with. On the fields I am most interested in as a lecturer and researcher—such as model-based historical composition, methodology of music theory, and analysis of music informed by scholarship from the time of its origin—AI does not have to offer much. Also, I don’t see any danger of obsolescence or redundancy of human creativity when it comes to producing acoustical ›recordings‹ of music inspired by traditional models of form, harmony, and voice-leading. In this respect, the results I have encountered so far are dull and uninspiring, and lack stylistic adequacy a great deal. While tools like Suno or Udio may yield an impressive outcome in some popular music styles, they desolately fail at tasks like generating, say, the beginning of a three-part fugue or a keyboard sonata exposition with two themes, modulating transition, and proper motivic treatment. See the first comment for one result of my painful attempts to have Suno produce the beginning of a classical piano sonata that moves from the I to V.

As long as generative AI models, prompted with text inputs, suffer from those obvious restrictions, they appear largely irrelevant for academic education in historically informed music theory. Still, I am curious what will happen in the near future. If we imagine that AI were able to generate something that sounds like an ideal-typical textbook example of a historical paradigm of musical form—say, a 16-measure small ternary, or a simple minuet with recapitulation of its initial idea—we might perhaps obtain something that could illustrate the function of a mere compositional model, plainly represented by the AI production for want of any existing musical embodiment. If we then confront this product with an actual piece of art that is based on the same model but transgresses the ideal-typical ›rules‹ in various aspects, and only then appears as an individual creation, such a comparison might in fact be beneficial as it could help students assess and better understand the role of formal and contrapuntal schemata for the genesis of a composition.

Preparatory Workshop in Music Theory

I will be offering a preparatory workshop again for music theory entrance exams at music universities in North Rhine Westphalia, organised by Deutscher Tonkünstlerverband NRW, and taking place in May 2024. Please feel free to notify your students who consider an application in Cologne, Detmold, Düsseldorf, Essen, or Münster!

» Thursday and Friday, 16–17 May 2024, 3–7 pm
» Online via Zoom | Participation fee: EUR 65

Some Thoughts on Absolute Pitch

When I was diagnosed with absolute pitch as a kid, I took it for granted and thought of it as no particularly remarkable capability. I was so used to be able to identify and produce pitches on demand that it astonished me to learn that others weren’t. Later during my adolescence, I more and more realised what a rare and freaky feature I had at my command. It kind of contributed to my nerdy reputation at school and was frequently mistaken for a proof of extraordinary musicianship or talent (which I was sure it wasn’t). Sometimes I even was annoyed by it, because my listening and musical experience was inextricably linked to my incorruptible pitch perception. It was just not possible to turn it off for a minute.

At university, when surrounded by a significantly higher percentage of people with similar abilities, I recognised the special challenges associated with switching between absolute and relative ways of aural perception. Demands such as listening to music in historical temperament, singing in a choir that could not keep tune, or sight-transposing sheet music posed considerable difficulties to me. I ended up in deliberately training my aural skills with respect to interrelations between separate pitches and complex harmonies, and to focus on structural aspects rather than individual elements in music. While doing so, my absolute pitch began to change and gradually became less reliable in certain contexts (especially with regard to unfamiliar sounds and instruments), but this was more than outweighed by a considerable gain in flexibility and listening strategies at hand. I also learned to adapt my ear and to perceive recordings of early music in the key in which they were meant to sound. A critical factor in this process was, so I believe, my own vocal practice in various choirs and ensembles.

During my studies, being in possession of perfect pitch was still relevant yet by no means crucial, and even though I passed every aural skills exam with distinction, I sometimes wondered what this capability was all about. It sure helped in my everyday work, providing quick and reliable tonal orientation while listening to long and complex pieces at all times, but it didn’t feel indispensable to me. After graduation, when I started to lecture in music theory and aural skills myself, it wasn’t so much a thing anymore. Students hardly ever asked me how to deal with absolute pitch, and I never felt inclined to have it served with any particular attention or methodology in my teaching. It is my conviction that relative aural skills need to be focused and cultivated as much as possible, while absolute skills, be they present or absent, do not. Consequently, I refuse to accept that perfect pitch should be of any relevance for the assessment of other crafts and musical proficiencies.

Analysing Form in Popular Music

I compiled an overview tutorial on the formal analysis of twentieth-century popular music, available on the platform Open Music Academy, and including the models of strophe and refrain, AABA form, verse-chorus form and its variants, as well as structural elements of EDM productions. Represented styles and genres are Broadway musical, rock’n’roll, rock, pop, and trance; examples are taken from the music of George Gershwin, Elvis Presley, Jimi Hendrix, The Beatles, ABBA, and Faithless so far. This is my first serious approach to examining popular music, so I’ll be grateful for any remarks and comments—in particular, if you have suggestions for including convenient examples from rap music, metal, or punk rock.

New Plugins on Open Music Academy

I am excited to announce that my digital teaching fellowship at Robert Schumann Hochschule Düsseldorf has yielded some results. The public platform Open Music Academy, currently run as a project of Hochschule für Musik und Theater München, now features four new plugins that allow persistent user entries within tutorials, introducing functions such as form fields, file upload, and a customisable whiteboard. These plugins are available and ready for use within virtual learning rooms and collaborative environments, and also include a commentary function for assessment and individual feedback. Please feel free to create an account and give it a try!

It is my hope that these plugins will be applied with benefit for designing modular tutorials, teaching materials, and online assignments. Many thanks to the developers, Andreas Helmberger and Adriana Luchian, for their proficient coding, to Ulrich Kaiser for his most valuable support, and to some of my colleagues from the RSH music theory faculty who supervised and tested the new features.