Foundation Models Reveal How Audio Effects Shape Musical Emotion
A recent investigation published on arXiv (2509.15151) examines the impact of audio effects such as reverberation, distortion, modulation, and dynamic range processing on emotional reactions to music. Utilizing foundation models—extensive neural networks trained on diverse data—the researchers assessed these influences, as these models capture intricate connections among musical elements, timbre, and emotional significance. Through probing techniques applied to deep learning embeddings, they revealed intricate, nonlinear interactions between specific audio effects and perceived emotions, while also testing the reliability of foundation audio models. This research fills a gap left by earlier studies that primarily concentrated on basic audio characteristics instead of the systematic effects of audio processing on emotions.
Key facts
- Study published on arXiv with ID 2509.15151
- Investigates audio effects: reverberation, distortion, modulation, dynamic range processing
- Uses foundation models (large-scale neural networks pretrained on multimodal data)
- Applies probing methods to deep learning embeddings
- Uncovers nonlinear relationships between audio FX and emotion
- Evaluates robustness of foundation audio models
- Addresses underexplored area of systematic impact of audio FX on emotion
- Prior studies focused on low-level audio features, not audio effects
Entities
Institutions
- arXiv