CAPABILITIES
What SHE can do.
The possibilities that shape a piece, what every piece carries, how a piece is recalled, and what a delivery is built around.
PROCEDURAL ART
What Is Procedural Art?
“The idea becomes a machine that makes the art.”
Procedural art predates trained models. In music it is called algorithmic composition: a person writes the rules and names the possibilities, and the system builds within those limits, so each result is emergent. Musicians wrote the rules in SHE. Choose the possibilities, or tell us the genres, moods, and lengths you need and we tune a profile for you; nobody types a prompt.
Structured Harmony Engine
Procedural composition
The rules, written down and inspectable
Musical profile, genre, key, progression, instrumentation, mood, tempo
Emergent: controlled chance, so no two pieces are alike
The whole piece, with the reasoning behind it
Stores the reasoning: Every musical choice is written down, and any piece comes back by its ID.
Runs on standard CPUs. No data center, no GPU queue.
Nothing is trained or served, so there is no inference cost per piece and no third-party model in the pipeline.
A fraction of the power a trained model needs to build or run.
ORIGINALITY
Verified Origin through Construction
Originality stems from the construction process rather than a later claim. People built every part: experts wrote the rules, not a trained model, and the engine logs every musical choice as it composes.
Theory based
Each note originates from theory taxonomies. The inputs do not use recordings or trained models.
Transparent process
The taxonomies, progressions, and the route from the chosen possibilities to the final piece can be inspected. The engine is built to be readable.
Recall by ID
With the same SHE version, the same musical profile and possibilities, and the same ID, the piece reproduces note for note.
Change detection
SHE replays a piece from its record and compares the outcomes. It detects changes in the engine, the profiles, the weights, or the taxonomies.
Original taxonomies
The taxonomies are created from theory and tradition instead of being transcribed from existing tracks. The engine and its pieces begin from a clean foundation.
POSSIBILITIES
Song Possibilities
Eight possibilities shape a piece, and the space inside them is endless. Genre, mode, key, and progression alone give over a million starting points, and from each one new harmonic possibilities emerge as moods, rhythm patterns, instruments, and voicings combine: musical diversity by design.
Musical Profile
The hand-tuned character behind a style: about 150 parameters across thirteen concerns, from swing and chord color to which instruments lead in which section. A profile can be named for a genre or for a feeling.
Jazz, blues, house, or a profile of your own
Genre
The musical tradition for the piece. Each genre inherits from its family and overrides only where it differs, so new styles are authored, not trained.
80 genres across 18 families, from House and Jazz to Ambient and Hip-Hop
Key
The primary note that anchors the composition.
Any of the 12 musical keys
Mode
The character of the piece. SHE aligns each note to the chosen mode.
All 11 modes, including Major, Minor, and Phrygian Dominant
Progression
A specific progression or one chosen by SHE to match the genre and mood. Each is written as harmonic roles, so the same progression realizes differently in every mode.
107 progressions, including I-V-vi-IV, ii-V-I, and the 12-bar blues
Instrumentation
Which instrument plays each role: melody, bass, chords, pad, drums. Every instrument carries its playable range and the genres it belongs in.
129 instruments, from vibraphone and fretless bass to bowed pad and jazz kit
Mood
The emotional quality, like brightness, intensity, or density. Moods combine, and each one shifts what the profile reaches for.
28 moods, including Driving, Melancholic, Upbeat, Epic, and Dreamy
Tempo
The speed in beats per minute, determining the feel and energy. Each profile carries its own range, so a tempo left unset lands where the style lives.
50–240 BPM across the profiles, from new-age to flamenco
PROFILES
A Genre Is a Profile, Not a Label
Behind each genre sits about 150 hand-tuned parameters across thirteen concerns. Every profile is a difference from one universal base: a family sets what its members share, and each member overrides only where it differs. Eighty genres across eighteen families ship today, and a new style is written as data rather than code, so a profile named for a feeling is as valid as one named for a genre.
The values were set by ear. Nearly four hundred notes in the profiles record why each value is what it is, including the alternatives that were heard and turned down.
- Rhythm and swing
- Harmony and chord color
- Voicing
- Phrase structure
- Energy by section
- Arrangement and form
- Articulation
- Rests and breathing
- Timing grid
- Transitions
- Dynamics
- Pattern choice
- Bass and drum lock
- Ambient fades in from silence and back out, holds wide three-voice chords, and never writes a pre-chorus: its transitions are swells.
- Rock plays power chords, root and fifth with the third left out, puts a lift before every chorus, and strips to the pad under the bridge.
- Jazz is head, solo, head out: the rhythm section never drops, the pianist lays out behind the soloist, and there is no pad at all.
OUTPUT COMPONENTS
Primary Records and Data
The structure is built first, and the audio is a rendering of that structure, the way a printed page is a rendering of a manuscript. Each piece produces an audio recording, structured metadata with MIDI mapping and musical descriptions, and a log of all musical decisions.
Every label is a decision the engine made: musical profile, genre, key, mode, progression, instrumentation, section form, mood, and tempo. Nothing is estimated from the audio afterward.
- Full song or separate tracks for chords, bass, melody, drums, and pads
- Uncompressed WAV files using standard sample rates
- A SHA-256 digest beside every file, so a delivered render can be verified
- Ready for playback, production, or integration
- Note-level instructions for each instrument
- Separate tracks for instrument roles, including velocity and timing data
- Compatible with music production software
- Musical profile, genre, key, mode, tempo, and time signature
- Progression, chord-by-chord harmony, and instrumentation for each section
- Section markers with bar counts and energy levels
- Instrument roles documented during the creation process
- A record of each musical decision and its rationale
- Content hashes of the data and parameters, and every refinement edit with its reason
- Machine-readable data paired with audio for labeling
- The progression, form, and ID needed to rebuild the piece
DELIVERY
Built Around What You Need the Music to Do
Every engagement starts from your goals: what the data will train or test, which labels matter, and how much coverage you need. The default package is shaped to fit, and what a given delivery includes is set by the terms of each engagement.
- Coverage
- Target the genres, modes, keys, tempos, progressions, and densities you need, drawn from 80 genres and 11 modes, including rare cases on demand.
- Volume and schedule
- A one-time corpus or recurring batches, scaled to the volume you need.
- Audio
- Uncompressed WAV renders, as a full mix and as separate parts for each role.
- MIDI arrangement
- One track per role: melody, bass, chords, pad, drums.
- Label metadata
- Musical profile, genre, key, mode, progression and chord-by-chord harmony, instrumentation, section form, mood, and tempo, as JSON or mapped to your taxonomy.
- Provenance record
- Every engine choice, content hashes, and the edit log, replayable note for note from its ID.
- Quality report
- Results of the eight-part automated gate for every song: pitch, rhythm, voicing, timing, groove, file integrity, melodic breathing, and progression fidelity.