Hi everybody!
I wrote my Bachelor's thesis under the supervision of Paul Preuschoff on SoundMuse: Contextualized Sound Cues for Creative Writing. The Work focuses on creative writing, a process that is often hindered by challenges such as concentration difficulties or writer’s block. Previous research has shown that music and ambient sounds can positively influence writing. However, such effects are not always consistent - sounds can be distracting or fail to set the desired mood. This raises the question of whether a soundscape dynamically tailored to the written content could improve the writing experience. To explore this idea, we developed the SoundMuse prototype, a system that uses AI to generate and play context-specific auditory cues in real time. By analyzing the evolving text, the system continuously extracts semantic and emotional cues and translates them into relevant audio feedback. The system uses OpenAI’s GPT-4 to detect such cues and generate corresponding tags, which are then used to retrieve fitting audio from existing databases. We conducted a small preliminary study to investigate how such adaptive soundscapes affect the writing process. Initial results suggest that if the audio transitions are smooth and thematically appropriate, such systems can promote immersion and creative flow. At the same time, the study underlines the importance of a considered design to avoid distractions.
At the chair, I also attended the M3 lab, which led to the following collaborative game project.