Structured Harmony Engine
SHE composes the music and the data describing it. Every note has a reason.
LEARN MOREWHAT SHE IS
SHE is a structured data synthesizer.
This procedural engine composes from music theory, renders an audio recording, and logs every musical decision, all in one pass.
Data in
- Genre
- Electronic
- Mood
- Upbeat
- Key
- D minor
- Tempo
- 137 BPM
Logic derived
- Progression
- vi – IV – I – V
- Parts
- Bass, chords, melody
- Phrasing
- Rests, then human timing
Music out
- Intro
- Verse
- Chorus
- Bridge
- Outro
Metadata out
- Description
- Every setting, named
- Decisions
- Each choice, and why
- Parts
- Editable MIDI
- ID
- Rebuilds this piece
Written, not trained. No artificial intelligence, no existing recordings, only a hand-authored library of musical theory and culture.
WHO BUILDS WITH SHE
Original at Scale
The music starts as a structure. The notes and composition rules are transparent, documented, and readable by humans. Although the music can surprise a listener, the engine knows precisely why each note was played.
Buyers with low risk tolerance need music at scale. This covers teams that train models on music, ship music within products, or score content at scale. Most existing options are scraped, estimated, or licensed under terms that forbid these particular uses.
Music Technology
Ground-truth data for music models.
Generative audio is constructed from existing material and lacks an underlying structural layer. This makes it difficult to know what the model learned and how to improve it.
Gaming
Music that evolves alongside the player.
Middleware rearranges assets that require licenses per title, which limits the music. Repetitive loops and short tracks are common, and the music can't adapt to the player or the game state.
Enterprise & Background Audio
On-brand music across any scale, live.
Catalogs and sample libraries possess a fixed size and use metered pricing. This makes it difficult to scale music for large teams or to provide a continuous stream of original music.
Music Education
Theory, heard and seen simultaneously.
Teachers can use the engine as a tool to teach how music is built and why it works, which helps students understand the underlying principles of music theory. The engine can also generate examples of musical concepts in real time.
SEE A COMPOSITION
Press Play
Audio
WAV · full song + 5 tracks
MIDI
Note-by-note and fully editable, one track per instrument.
A structured description
- Key:
- D
- Mode:
- Aeolian
- Tempo:
- 137 BPM
- Meter:
- 4/4
- Progression:
- vi – IV – I – V, realized as VI – iv – i – v
- Sections:
- Intro, verse, chorus, verse 2, bridge, outro
- Tracks:
- Chords, bass, melody, drums, pad
Produced while SHE builds it, not labeled afterward.
A complete record
- ID:
- 53fb33d31684728e
- Progression:
- Template vi – IV – I – V, realized in aeolian as VI – iv – i – v
- Mood:
- Driving, groovy at 137 BPM
- Energy:
- Intro ▁ verse ▃ chorus ▇
- Parts:
- Bass on roots, melody steps to the next chord
Every musical decision SHE made. Replay the ID to rebuild it exactly.
HOW IT COMPOSES
From Theory to Audio
The engine chooses a style and mood, determines the required musical parameters, and constructs the composition within those constraints. Each decision sets the foundation for the next one. The final pieces's record contains the logic that was used to create it.
Theory
The foundation: how music works. Intervals, harmony, rhythm, and meter, true across every style.
Library
The collections: chord patterns, rhythms, styles, moods, and instruments, each one drawn from tradition, validated, and organized.
Orchestration
Where the music takes shape: choosing the chords, shaping the bass, writing the melody, planning the sections, and placing the rests so the song can breathe, making decisions the way a player does.
Export
Putting it together and handing it off: the audio, the editable parts, the description, and the record, all at once.
LIVE OUTPUT
NEWLive Stream
SHE can also stream content. It creates a nonstop playlist of original songs, each track being unique. This service is designed for retail settings, video games, and broadcast backgrounds that require uninterrupted music.
NOTHING WAS TRAINED
No Model, No Artist Recordings
Generative models depend on billions of parameters, the original recordings used for training, and substantial hardware. This engine does not need those resources because it never underwent a training process.
The engine doesn't use material from other sources. Each piece is generated wholly inside the engine, which keeps a full record of its origin.
Training or serving is not required. This eliminates inference costs per piece, GPU queues, and the necessity for third-party models in the pipeline.
The system operates on standard low-power CPUs and does not need a data center.
Energy consumption is minimal compared to the power needed to train a model.