< Back to IRCAM Forum

Praat AudioTools

Artistic Research
#1

Dear colleagues,

I would like to share a new project on the IRCAM Forum:

Praat AudioTools

Praat AudioTools is a collection of scripts designed to extend Praat’s functionality for audio analysis, structured processing, and batch-oriented workflows. It aims to support research applications and experimental sound practices.

Feedback on functionality, performance, and possible extensions is very welcome.

Best regards,
Shai Cohen

#2

Update – Hybrid Systems

I have recently added a new section titled Hybrid Systems, integrating external Python processes into the compositional workflow.

From a composer’s perspective, this opens the door to working with deep learning and AI models as part of the creative process. Analysis and resynthesis remain anchored in Praat, while neural networks and machine learning systems can reinterpret, transform, or abstract the material before it returns as sound.

It creates a space where acoustic structure meets artificial intelligence — not as a replacement for compositional thinking, but as an extension of it.

I warmly invite anyone interested in AI-assisted composition and hybrid workflows to explore the Hybrid Systems section and share thoughts.

Best regards,
Shai Cohen

#3

Praat AudioTools can run from MAX/MSP

#4

Dear colleagues,

I would like to share a substantial expansion of the Hybrid Systems section in Praat AudioTools.

These additions continue the same general direction: keeping analysis and resynthesis anchored in Praat, while extending the workflow through external Python engines for machine learning, latent-space navigation, granular recomposition, spectral transformation, multichannel rendering, and interactive control.

Some of the new additions include:

AI / latent-space composition
AI Conductor Mix, Hierarchical Neural Recomposition, Identity Separation, Latent Barycentric Mutation, Latent Counterpoint, Latent Diffusion / Morph-Chain Generator, Latent Folding, Latent-Event Relocation, Latent Space Navigation, Latent Spat, Latent STFT Decoder, Self-Attention Latent Navigation, Temporal Elasticity, and Thermodynamic Transform.

Corpus / granular / phrase-based recomposition
Corpus Mosaic, Formant Swarm Granulator, Granular Navigation Engine, Phrase Rewriter, Phase-Space Composer, Recomposer, SSM Morph Composer, and Rhythmic Voice Flattener.

Spectral / temporal transformation
HPSS Phase Vocoder, Internal Polyphony, Phase Diffusion Engine, Spectral Eraser, Spectral Morph, and Sympathetic Resonance.

Spatial / control / integration tools
Envelope Editor, Multichannel Playback Bridge, Praat Pbind, Praat for Max and M4L, Self-Reflective Feedback, Spatial Panner, TinySOL Orchestration Retrieval, and VST3 Effect.

A few examples:

  • AI Conductor Mix assigns dynamic roles and states to segments of multiple sounds and renders a stereo mix plan in Praat.
  • Corpus Mosaic rebuilds a target sound from a corpus using grain-level feature matching, with controls for repetition, continuity, and randomness.
  • Internal Polyphony uses NMF-based decomposition to reveal simultaneous internal voices already latent in the source.
  • HPSS Phase Vocoder performs time-stretching while preserving transient clarity by separating harmonic and percussive components.
  • Thermodynamic Transform analyzes acoustic structure, discovers phase-like regimes in the material, and applies regime-dependent transformations to individual audio events.

The new additions also extend the system outward: TinySOL Orchestration Retrieval brings orchestral sample matching into the workflow, while Praat for Max and M4L connects Praat AudioTools with Max/MSP and Ableton Live for hybrid studio practice.

The broader aim is not to replace compositional thinking with AI, but to open a modular space where acoustic analysis, machine learning, formal transformation, and compositional decision-making can interact inside a practical Praat-centered workflow.

I would be very glad to hear thoughts, questions, or suggestions—especially from people interested in AI-assisted composition, research-creation, hybrid audio workflows, or experimental sound transformation.

Best regards,
Shai Cohen
https://mashav.com/sha/Praat%20AudioTools/script-overview.html