Abstract Praat AudioTools is an open-source library of over 300 scripts that transforms the phonetic analysis software Praat (Boersma and Weenink) into an offline, object-centric laboratory for electroacoustic composition. While Praat is the standard tool for linguistic analysis, its potential for musical creation has remained largely untapped due to its interface design. We propose a workflow in which phonetic analysis objects – specifically PitchTiers (frequency curves), FormantGrids (resonance tracks) and TextGrids (temporal segmentation) – function as editable musical scores. Unlike real-time performance environments (e.g., Max/MSP, SuperCollider), which prioritise low-latency interaction, this toolkit emphasises ‘compositional deep time’, embedding analysis within an iterative edit–render–listen loop. Small modifications to analysis data produce structural consequences in timbre, gesture and form, enabling a research-creation practice rooted in the acousmatic tradition and spectromorphological thinking. By treating phonetic measurement as compositional material, AudioTools bridges phonetics and poetics. We contextualise this framework within the lineage of speech synthesis in electroacoustic music and demonstrate, through case studies, how it enables compositional strategies grounded in analysis as composition. The toolkit integrates neural network processing while maintaining interpretability, positioning it critically against black-box neural synthesis and arguing for transparent, parametric control in research-creation practice.
S. Robin Cohen (2026) studied this question.