The Doorway Between Beats and Seconds
08 November 2026
Methods essay

Abstract
A musical event has more than one address. A composer may place it in a bar or phrase; a performer realizes it through timing and gesture; a digital renderer stores it at sample indices. Praxymuse v0.6 makes these layers explicit through tempo mappings, embodied-event records and long-range structure. This essay examines their interfaces rather than repeating the earlier plan-fidelity experiment. Its central question is how a representation can preserve musical relationships while translating between different clocks. [1, pp. 2–5]

Three clocks, one event
The source distinguishes audio time s, beats b and integer sample index n. For positive tempo B(s), it defines b(s)=∫₀ˢ B(u)/60 du. Positivity permits an ordered beat map under suitable regularity. An event scheduled at beat position b_j with an expressive displacement δ_j is mapped back to seconds through s_j=b⁻¹(b_j+δ_j), then to a sample by n_j=round(F_s s_j). [1, p. 4, M04]

The rounding bound |n_j/F_s−s_j|≤1/(2F_s) expresses a limited numerical guarantee. It bounds conversion to the nearest sample at sample rate F_s. It does not establish good groove, faithful performance or perceptual indistinguishability. Those depend on the musical context and on how event timing interacts with articulation, dynamics and other parts.

Where a simple grid becomes insufficient
A score grid can locate attacks while leaving their relationships underspecified. Praxymuse models groove through joint timing and velocity information, and its embodied-event layer conditions phrase variation on score, skill, body, affect and neighboring events. The proposed covariance is contextual rather than automatically diagonal. [1, pp. 3–4, M02 and M04]

That distinction matters because independent random jitter is not the same as coordinated performance. Two notes may move together to preserve a gesture; a bass answer may intentionally lag a lead; a breath can organize the next phrase. These are possible design relationships, not acoustically measured effects claimed here. A renderer needs controls that can realize them, and an analyst needs observations before asserting that they occurred.

The interface must remember the instrument
M02 represents an instrument through source, excitation, resonance, range, tuning, polyphony, envelope, controls, limits and articulations, with culture and provenance kept as fields. It then gives feasibility conditions for events such as range, reach, breath and technique transitions. These are proposed modeling contracts, not proof that all named instruments have been implemented. [1, p. 3]

A phrase translated from a keyboard plan to a wind instrument cannot be preserved merely by copying pitches and timestamps. Breath and transition constraints may change the realizable event sequence. The translation should therefore preserve the intended musical role while recording necessary changes, rather than labeling a mechanically copied sequence an authentic performance. The architecture's value lies in exposing such decisions instead of hiding them inside a generic instrument name.

Structure above the event
The composition graph in M07 connects sections, phrases, motifs, lyrics, affect arcs, harmony, tempo, meter, roles and transitions. Its proposed factorization moves from form and arc to motifs, prosody, ensemble, gestures and audio. The source explicitly treats that factorization as a model with conditional-independence assumptions, not a demonstrated generative law. [1, p. 5]

This gives the timing interface a larger purpose. A local delay can support a phrase, but its meaning also depends on where the phrase sits in the form. A return of a motif after a contrasting section can be recognizable without being an identical event list. A useful representation distinguishes structural identity from literal repetition, allowing transformations while preserving the declared relationship.

A testable translation exercise
A small study could begin with an original phrase at constant tempo, then introduce a continuous tempo change while preserving beat positions. Check monotonicity of the beat map, numerical inversion error and sample-rounding bounds. Add a declared breath constraint and compare the resulting event plans before any synthesis. Finally, render with a documented backend and evaluate whether listeners recognize the intended gesture. This is a proposed sequence, not an experiment performed for this essay.

The measures should remain separate. A numerically accurate beat-to-sample conversion does not establish expressive success. A preferred performance does not prove that every symbolic constraint survived. The source's separation of operational controls, technical verification and listening judgments supports this layered interpretation. [1, p. 8, M15]

Conclusion
The most important architecture may be the boundary between representations rather than the largest model inside them. Beats preserve musical placement; seconds coordinate performance; samples enable digital playback. Instrument and phrase records explain why conversion may require more than arithmetic. Praxymuse offers a vocabulary for making those joints explicit, while leaving their implementation and musical adequacy open to testing.

Reference
[1] Praxymuse Theorem v0.6: Source-preserving mathematical integration, 6 November 2026, original PDF pp. 2–5 and 8, modules M02, M04, M07 and M15. Source: Praxymuse_Theorem_v0.6_Full(2).pdf. This essay addresses the specified public version and does not claim it is the latest release.
