Abstract: The Audio Metaphor (AUME) demonstration invites participants to explore an interactive system for automatic soundscape composition. AUME is built as an online prompt-based system which produces audio soundscapes given a textual and affective query. The user can set valence and arousal curves which the system uses to generate a soundscape composition matching the desired eventfulness and mood. By streamlining the composition process, AUME aims to relieve some of the cognitive work done by sound designers. AUME's latest iteration draws from a database of nearly half a million audio files, each paired with a list of textual sound descriptors. Along with this expanded capacity, comes a refined ability to interpret a wider lexicon describing sound. Improvements to AUME’s algorithms and architecture for sound retrieval, segmentation, background and foreground classification, automatic mixing and automatic soundscape affect recognition, makes it a powerful system that generates believable soundscapes at interactive rates.