People
People » Graduates and Alumni » Research Students » Huw Davies, Worcester College
Huw Davies, Worcester College
Research Project: Innovation in Dynamic Music for Video-Games - Branching and Generative Musical Engine Prototypes
Like music for film and for the stage, video-game music seeks to mimic or accompany the actions or emotions shown on screen. However; unlike these other forms of media, video-game music is unique. It exists without the normal time conditions or time constraints attributed to these static media. This makes the music in video-games dynamic or active. It is active in the sense that it must react, however crudely, to the situations occurring because of the direction of the player. This activity is controlled by what is effectively a musical artificial intelligence. The relative youth of the video-game industry is one factor behind the lack of research into the development of these musical artificial intelligences (or sound engines) and amplifies this project’s future impact.
To give an example of the role of a sound engine: in The Legend of Zelda series when Link (the player’s character) is in a village, music associated with the village will be played (‘village’ music); when link enters a shop, within the village, different music will be played (‘shop’ music). This simple example unearths a large problem with current sound engine design. While Link is outside the shop there must be accompanying music; however, because the player has complete control over whenLink moves into the shop the music must be written to last for an indefinite amount of time. Currently composers use small sections of music and will loop them to maintain a constant backing track to the player’s actions but this creates the problem of abrupt changes in musical texture/harmony/melody and aesthetic.
My project seeks to design new dynamic music engines that cope with two different situations where musical aesthetic is currently not upheld within video-game music. It is multidisciplinary in the sense that it requires knowledge of aesthetic musical concepts, aesthetic video-game concepts and an ability to create active musical engines by way of program building.
The first engine I discuss achieves dynamic transitioning between two musics with a case study of a ‘battle sequence’ from the Final Fantasy series: Even up until current games in the series, within this scenario the looping music of the ‘battle’ is abruptly ended when the player-controlled party achieves victory. My goal was to create a program and the corresponding music that had the ability to intelligently select a branching musical pathway effectively linking one area of music (the looping ‘battle’ music) to another (the ‘victory’ music), crucially, in an aesthetically consistent way. To do this I needed to create a program that understood its position within a larger piece of looping music. Once this was achieved I could then compose numerous different potential bridging passages from the one music to the other. When activated the program would read where it is within the loop and play the corresponding bridge passage to the new music. This engine for dynamic musical branching can be applied to many other situations beside the specific ‘battle sequence’ in the case study.
I am now focusing my work on a second, much more complex engine, aiming to create a program that can generate music on-the-fly for any emotional or dramatic situation. This entails building five linked hierarchical levels of control, each one adding a further level of intelligence to the system. The lower levels are randomly generated with the higher levels constricting that randomness into a perceived musical language. The musical detail, such as individual note lengths and pitches are highly randomized; however, when put under the rigorous constraints of the higher functions will average out to a consistent and predictable musical style. This type of musical system is analogous to the scientific prediction of chaotic systems. Though components at the microscopic level are impossible to predict, the macroscopic music produced is predictable and results in applicable music for the given situation.
This generative musical engine lends itself to more so-called ‘openworld’ exploratory games and perfectly maintains the transitional aesthetic consistency that the branching-music engine above worked towards. The separate strands of this project push the boundaries of the interactive entertainment experience by using a combined scientific and artistic approach. By attempting to produce aesthetic music through artificial means this project takes some of the first steps towards creating a new type of musical artificial intelligence, drastically increasing the music’s connection with the onscreen action and heightening the immersive experience for the player(s).