Tranen op de Dansvloer: a tearjerker by A.I. (2020)

Project for SETUP.

Blood, sweat, and tears! Virtual ones, that is.

Sung by an impassioned singer and provided with a real sob, the Dutch "tranentrekker" or "smartlap" (tearjerker) is the epitome of human emotion. Now that crying in the pub is out of the question, we turn to our new best friend: the computer.

Can we train algorithms to capture the typical sound and heart-felt sentiment so present in these songs? What does it entail to compose music in collaboration with AI? And what future role do we envision for AI in the creative process?

In collaboration with AI researcher and musician Janne Spijkervet, algorithms GPT-2 and Music Transformer, music producer Manfred Jongenelis and the Utrecht based "smartlappen" band Pure Ellende, we researched how to create a "smartlap" with artificial intelligence.

The result is a song co-created by humans and machines. Watch the videoclip here!

I guided the artistic research, and produced, hosted and edited a series of podcasts in which we reflect on the process. How did we go about training the models? What was selected as its input? How did the musicians experience co-creating together with the algorithms? What do they think about the algorithms' role in the creative process, and what are their thoughts on how this will impact making music in the future?

Listen to the talks here. [Dutch]

1/3 The preparation: how to explain an algorithm what a "smartlap" is?

2/3 The creation: how to write lyrics and compose music with AI models?

3/3 The result: how does it sound? Reflections for the future.

More information about the creative process and  results at

© Bas van Setten