Spectralism reimagined: AI, orchestration and improvisation

This article examines how algorithms are embedded in contemporary instrumental and electroacoustic practice, focusing on orchestration tools and AI-improvisation environments such as Orchidea and Somax2. Its central claim is that artificial intelligence in these systems extends spectralist compositional tendencies into new computational and performative contexts. Historically, instrumental resynthesis recreated analysed spectra through vocal-instrumental, or electronic means; today, AI-based frameworks broaden this logic, opening new aesthetics and modes of interaction. In both systems, spectral analysis underpins the segmentation, combination and transformation of sound within corpus-based synthesis. Orchidea maps spectral targets onto instrumental resources through forecasting and optimisation, enabling new approaches to orchestral design. Somax2 navigates its corpus in real time through multi-agent interaction with human performers, reframing improvisation as a dynamic feedback loop.