Methodology Notes
Expose the methods, calibration, and safeguards that make the institute credible.
Methodology notes explain how Zeltor Labs assembles the corpus, calibrates benchmarks, validates inferences, and keeps the public pages inspectable. This is where the lab earns trust under scrutiny.
Stream markers
Methodology and evidence
Methods, safeguards, and evidence-chain notes.
MN-01
Corpus assembly
March 2026
8 min
How Zeltor Builds the Reference Corpus
A methodology note on how the MusicBrainz-based reference field is assembled, stratified, and used without pretending scale alone proves quality.
Read noteMN-02
Source discipline
March 2026
8 min
Citation Chain and Evidence Standards
A methodology note on how Zeltor Labs links published literature to public benchmark and research claims.
Read noteMN-03
Threshold setting
March 2026
8 min
Benchmark Calibration and Reference Bands
A methodology note on how benchmark ranges become usable public corridors rather than arbitrary labels.
Read noteMN-04
Client-side processing
March 2026
8 min
Browser Audio Analysis Without Uploads
A methodology note on how MusicMind performs public analysis while keeping audio in the browser.
Read noteMN-05
Tonal inference
March 2026
8 min
Harmonic Inference and Key Confidence
A methodology note on how tonal center, mode, and ambiguity are inferred and surfaced publicly.
Read noteMN-06
Structural inference
March 2026
8 min
Section Boundary Detection Explained
A methodology note on how section changes, labels, and form types become public structure reads.
Read noteMN-07
Affective inference
March 2026
8 min
Emotion Arc Modeling and Validation
A methodology note on how emotional movement is modeled without pretending mood is one-dimensional or objective.
Read noteMN-08
Sonic posture features
March 2026
8 min
The Timbral Fingerprint Layer
A methodology note on how brightness, warmth, contrast, and texture become a public timbral fingerprint.
Read noteMN-09
Groove feature design
March 2026
8 min
The Rhythmic Genome Layer
A methodology note on how groove, density, syncopation, and movement quality are organized into a readable rhythm layer.
Read noteMN-10
Pressure over time
March 2026
8 min
Dynamic Contour Modeling
A methodology note on how section spread and arc spread are used to describe dynamic contour publicly.
Read noteMN-11
Playback context
March 2026
8 min
Translation and Playback Fitness
A methodology note on how playback contexts influence benchmark interpretation and release fitness.
Read noteMN-12
Audio-to-visual mapping
March 2026
8 min
Visual Correspondence Inference
A methodology note on how music analysis is translated into visual direction and generative scene behavior.
Read noteMN-13
Public explainability
March 2026
8 min
Interpretability and Human-Readable Outputs
A methodology note on how technical analysis becomes public language without collapsing the underlying nuance.
Read noteMN-14
Statistical hygiene
March 2026
8 min
Outlier Handling and Distribution Cuts
A methodology note on how outliers and edge cases are handled so public distributions stay useful.
Read noteMN-15
Data stability
March 2026
8 min
Schema Versioning for Music Intelligence
A methodology note on how public data structures and release pages stay stable as the lab evolves.
Read noteMN-16
Decision-layer safety
March 2026
8 min
Release Scoring Guardrails
A methodology note on how benchmark pages and the report are kept from collapsing into simplistic score worship.
Read noteRelated publication paths
Method notes explain how claims are built and checked.
Methodology notes show how the corpus, benchmarks, and browser-native analysis become public claims without overclaiming what the system can do.