Automatic Generation of Dynamic Musical Transitions in Computer Games

Cutajar, Simon (2020). Automatic Generation of Dynamic Musical Transitions in Computer Games. PhD thesis The Open University.



In video games, music must often change quickly from one piece to another due to player interaction, such as when moving between different areas. This quick change in music can often sound jarring if the two pieces are very different from each other. Several transition techniques have been used in industry such as the abrupt cut transition, crossfading, horizontal resequencing and vertical reorchestration among others. However, while several claims are made about their effectiveness (or lack thereof), none of these have been experimentally tested.

To investigate how effective each transition technique is, this dissertation empirically evaluates each technique in a study informed by music psychology. This is done based on several features identified as being important for successful transitions. The obtained results led to a novel approach to musical transitions in video games by investigating the use of a multiple viewpoint system, with viewpoints being modelled using Markov models. This algorithm allowed the seamless generation of music that could serve as a transition between two composed pieces of music. While transitions in games normally occur over a zone boundary, the algorithm presented in this dissertation takes place over a transition region, giving the generated music enough time to transition.

This novel approach was evaluated in a bespoke video game environment, where participants navigated through several pairs of different game environments and rated the resulting musical transitions. The results indicate that the generated transitions perform as well as crossfading, a technique commonly used in the industry. Since crossfading is not always appropriate, being able to use generated transitions gives composers another tool in their toolbox. Furthermore, the principled approach taken opens up avenues for further research.

Viewing alternatives

Download history


Public Attention

Altmetrics from Altmetric

Number of Citations

Citations from Dimensions

Item Actions