Notation
Specification

An MNN string is a single line of text composed of tagged segments. Each segment is wrapped in brackets and prefixed with a tag identifier. The segments can appear in any order — parsers identify them by tag, not position. A minimal MNN string needs only a movement tag and a contraction tag. All other segments are optional and additive.

The notation is designed to be human-authored. Unlike motion capture formats (BVH, C3D) that are machine-generated time-series data, MNN is written by a person describing what they did, what they felt, or what they want a system to reproduce. It is closer to musical notation than to a data dump.

The contraction segment identifies which muscles are active and how intensely. Muscles are referenced by abbreviated anatomical names. Activation levels use a + scale where more symbols mean greater engagement. The primary mover is listed first, synergists follow.

Contraction types are implicit in the movement tag but can be made explicit: concentric (muscle shortening under load), eccentric (muscle lengthening under load), isometric (static hold), and ballistic (explosive). This matters for rehabilitation logging where the distinction between concentric and eccentric loading of a tendon is clinically significant.

The nerve segment records which peripheral nerves are responsible for the movement. This is optional for general fitness logging but critical for rehabilitation and clinical documentation. A physical therapist tracking a patient with a brachial plexus injury needs to know which nerve pathways are being loaded, not just which muscles are moving.

Nerve annotations also enable tracking of nerve flare-ups alongside sets. If a user reports radiculopathy or paresthesia during a specific movement, the MNN string captures both the movement pattern and the affected nerve — creating a record that a clinician can use for treatment planning.

The position segment describes the orientation of joints during the movement. Joints are referenced by abbreviated names, and rotations are expressed in degrees using standard anatomical terms: flexion/extension, abduction/adduction, internal rotation/external rotation.

This is the layer that makes MNN work across domains. The same position data that a trainer uses to describe a cable fly angle is the data that InThisWorld uses to pose an avatar. Position values map directly to Euler rotations in a 3D engine — the translation from [Pos:L.Sh(IR:25,Flex:90)] to llEuler2Rot() is mechanical, not interpretive.

The vector segment describes where the resistance comes from, at what height, and at what angle. This is the layer that makes MNN work for cable rigs, robotic rehabilitation, and remote-controlled equipment. A BODWAVE-compatible cable system can read a vector segment and set its pulley to the correct height and angle to reproduce the prescribed movement pattern.

The distinction between a high cable fly, a mid cable fly, and a low cable fly isn't just about the name — it's about the line of resistance relative to the muscle's force vector. The vector segment captures this precisely. A system that knows the height, angle, and source can reproduce the resistance profile on any compatible equipment.

MNN is consumed by three AIUNITES sites, each using a different subset of the notation:

Existing standards each capture one layer of movement data. MNN combines all of them into a single string. This is not a criticism of prior work — each system was designed for a specific domain. MNN's contribution is the integration layer that makes movement data portable across domains.

VRN describes voice production through the physical configuration of the vocal tract. Each VRN record is a set of parameters representing the state of the larynx, vocal folds, tongue, jaw, lips, soft palate, and breath system at a given moment. Together they define a complete vocal gesture — everything the body is doing to produce a specific sound.

The parameters are grouped into three categories: source (what the vocal folds are doing), filter (how the vocal tract shapes the sound), and drive (the breath energy powering it). This maps to the source-filter model of speech production — the same theoretical framework used in acoustic phonetics and speech synthesis research.

Current AI voice synthesis engines — ElevenLabs, OpenAI TTS, Google Cloud TTS, Amazon Polly — all model voice as an acoustic waveform transformation. They take a text input, select or clone a voice identity, and produce audio. Some accept style parameters (speed, emphasis, emotion). None of them model how the voice is physically produced.

This means these systems can produce convincing speech, but they cannot answer questions like: "What is the larynx doing during this vowel?" or "What tongue position produces this formant shift?" The physical reality of vocal production is invisible to them — it was never part of their parameter space.

VRN fills this gap. It provides a parameter space that maps to the physiological mechanics of vocalization. A VRN record doesn't describe what a voice sounds like — it describes what the body is doing to produce that sound. This is a fundamentally different representation, and it opens a different research direction: physically-informed voice synthesis where a model generates speech by simulating vocal tract configuration rather than pattern-matching spectrograms.

For vocal training (the current primary use case on VoiceStry), VRN enables instruction that goes beyond "make it sound like this." A vocal coach can describe the exact physical configuration that produces a target sound — larynx low, soft palate raised, tongue body forward, specific breath pressure — and the student can work toward reproducing that configuration, not just imitating the acoustic output.

VRN is consumed by two AIUNITES sites:

Traditional vocal pedagogy tells you what to do: open the throat, lift the palate, support from the diaphragm. VRN gives you symbols for these mechanical events. VNN adds the deeper layer — the nervous system that actually executes those commands.

Every VRN symbol implies a VNN command chain. When a score says [Zy] (zygomatic lift), the facial nerve (CN VII) is sending signals to the zygomaticus major. When it says [D+++] (maximum diaphragm), the phrenic nerve (C3–C5) is driving the primary breathing muscle. VRN is the score. VNN is the orchestra.

Six cranial nerves form the core wiring harness of the vocal instrument. Every VRN symbol traces back to one or more of these nerves.

The Vagus Nerve (CN X) deserves special attention. Its two critical branches divide the work of phonation: the recurrent laryngeal nerve (RLN) controls 4 of 5 intrinsic laryngeal muscles (fold adduction, abduction, body tension, closure), while the superior laryngeal nerve (SLN) controls the cricothyroid — the pitch muscle that stretches and thins the folds. Every register transition involves a shift in RLN vs SLN dominance.

Every VRN symbol traces to a specific nerve or nerve combination. VNN organizes these mappings into tiers based on who needs them: singers (Level 1), teachers (Level 2), and clinicians (Level 3).

Vocal production depends on muscle pairs that oppose each other. The balance point between them determines the sound. This is the vocal equivalent of MNN's agonist/antagonist pairing — and it's where most vocal problems live. VNN uses a ratio notation to express the balance between antagonist pairs.

For ENTs, speech-language pathologists, and vocal therapists. These codes document pathology, track recovery, and identify compensation patterns — the vocal equivalent of MNN's clinical layer.

The spec above covers the core VNN framework. The full VNN reference — including all Level 2 articulatory position codes (tongue, larynx height, jaw, palate), the complete vibrato/squillo/timbral color mappings, the emotional brain pathway (periaqueductal gray), vagus nerve and wellbeing research, and VNN as a neural training framework — lives on VoiceStry:

Read the full VNN reference on VoiceStry →

MNN and VRN share the same design philosophy. These aren't accidental similarities — they're deliberate constraints that make both notations work across contexts.