01
AI / ML
Machine learning
A model learns patterns from training material and uses those learned relationships inside the finished work—generating, transforming, recognising or organising sound.
Wired Moth · Field guideVersion 1.0Say what the machine actually does
Process, not hype
A small dictionary
of machine-shaped music
“AI music” is too blunt to describe a field this varied. Wired Moth names the dominant process, separates learning systems from ordinary algorithms, and treats the artist’s own account as evidence—not the way a record happens to sound.
All machine learning is algorithmic. Not all algorithms learn. A modular patch, a live-coded rhythm and a neural voice model may all be machine-shaped music, but they are not the same practice.
02
AI / ML
Neural audio
Machine-learning models operate directly on audio: synthesising voices or instruments, transferring timbre, creating latent variations or transforming a live signal.
03
AI / ML
Machine listening
A system analyses incoming sound or gesture—such as melody, rhythm, phrasing or timbre—and uses that interpretation to affect a performance or composition.
04
Not necessarily AI
Generative software
Software is configured to produce an evolving musical result within rules, probabilities, feedback paths or modulation relationships chosen by the artist.
05
Not necessarily AI
Generative hardware
Synthesizers, sequencers and effects are connected so that a musical system evolves partly on its own. The artist designs, starts, steers, edits or performs with that system.
06
Not necessarily AI
Live coding
The artist writes or changes code during performance, making musical structure visible and mutable in real time.
07
Not necessarily AI
Algorithmic composition
Formal procedures generate or organise musical material. These might use mathematics, probability, data, recursion, grammars or custom software.
08
Not necessarily AI
Rule-based composition
A score or system sets constraints within which performers, software or instruments make the work. Human interpretation can remain central.
What gets in
Receipts before taxonomy
We count
An artist, label, technical, festival or reputable editorial source connecting the stated process to the specific work—and evidence that it materially shapes the sound, structure or performance.
We do not count
Tracks generated wholly or substantially from prompts with no documented, meaningful human musical construction; AI used only for artwork or marketing; an unexplained Bandcamp tag; vague claims about an artist’s wider practice; or a classification inferred from sonic resemblance alone.
The red line
A prompt is not a process note
Wired Moth is interested in artists building instruments, systems, models, rules and relationships—not uploading the first complete song returned by a generator. Prompting can appear inside a larger human-led practice, but inclusion requires evidence of consequential composition, performance, editing, model-building, sound design or system authorship beyond choosing an output.