Musique · music notation and audio · Not generally available today
One owned music suite, built in the open — a notation engraver for writing scores and SONUS, an audio workstation for recording and mixing.
Musique is two tools that share one home: a notation engraver, where you enter notes and it sets them into engraved, print-ready scores and parts, and SONUS, a multitrack audio workstation, where you record, edit, and mix sound. The name is the plain French word for music, chosen because that is exactly what the suite is for. Musique is a tool a person writes and records their own music with. It does not compose for you, and there is no AI-generated music here.
Musique is not generally available today. No account is provisioned from this page, there is nothing here to download, and nothing here takes money. What you can do today is write to us and say what you need; this page promises no date.
What Musique is
Writing music and recording music are two different crafts, and most people who make music end up doing both. You write a part on a staff, and you also want to hear it played, layered, and mixed. Musique keeps both crafts under one roof so the score you engrave and the audio you record are not two disconnected worlds.
The notation engraver is the writing side. You place notes on staves — pitches, rhythms, dynamics, articulations, lyrics — and the engraver handles the typesetting: spacing, beaming, page layout, and the extraction of individual instrument parts from a full score. The goal is engraved output a musician can read from and a printer can produce cleanly.
SONUS is the recording side: a multitrack audio workstation. You capture audio and MIDI onto tracks, arrange and edit them on a timeline, and mix with levels, panning, and effects. It is the studio surface for turning a written idea, or a live performance, into a finished recording.
Musique is not generally available today. No account is provisioned from this page, there is nothing here to download, and nothing here takes money. What follows is the suite as it stands in code right now. Where a card describes something that runs, it names the file, the format, or the keystroke behind it. Where a card describes something we have not built, it says so and names what is missing. There is no generative or AI-composed music anywhere in it: you write the notes, the tool sets them.
The notation engraver: note entry to engraved score
The writing side works on one principle: you think about the music, the program handles the engraving. Here is what the notation code actually does today — each card with the part that runs and the part that is still gated, side by side.
Note entry: one spine for the mouse and the keyboard
From a focused staff, A to G add a note of that step, R adds a rest, 1 to 6 set the duration from whole down to a 32nd, . cycles augmentation dots 0 to 1 to 2 and back, and + and - walk the accidental from double flat to double sharp. Arrow keys move the selection and raise or lower a note by a diatonic step, with Shift for a whole octave. The toolbar buttons call the same functions the keys do, so there is no mouse-only route to any edit, and every operation returns a new score rather than mutating the one on screen.
Engraving math in integers, on a real page
The layout call returns a whole page measured in centipoints, at 100 centipoints to the point: 61,200 by 79,200 is US Letter, 612 by 792 points. Glyph and staff-line coordinates are integers on that page. The interactive editor converts that page space into panel space rather than re-deriving positions of its own, so the staff you edit on is drawn from the same integer positions the SVG and the PDF are drawn from.
Every glyph carries a stable, semantic id
The engine tags each glyph with an id of the form p{part}m{measure}v{voice}e{event} — p0m0v0e0 for the first event — unique within a score and stable across re-layouts of the same score. The editor binds selection and its accessibility tree back onto its own note ids through that semantic id rather than by position, so if the engine’s ordering ever drifts the join fails closed on the individual glyph instead of silently misassigning notes across voices. Glyph labels are real pitch strings such as C5, or the literal rest.
Score and parts: a part is a staff, and the controls are separate
Each part carries three independent controls. Show and hide is purely visual — it filters engraved glyphs and staff lines by staff id, and a hidden part can still sound. Mute and solo are audio, and they compose with each other rather than overriding. The honest limit, written into the model: these are staff-level only. The score model has no addressable part below a staff, so independent voices inside one staff share that staff’s part; four-part vocal writing works because each voice gets its own staff. A per-voice part surface waits on a score model that promotes voices to parts.
An on-ramp from MusicXML, with the gaps counted rather than hidden
There is a working import parser for MusicXML, the open published interchange format, mapping it onto the score model: staves, measures, notes, pitches, durations, time signatures, clefs. It ships its own dependency-free XML reader, so it runs server-side and under the test runner without a browser DOM. Everything the model does not represent today — grace notes, tuplets, ties, lyrics, unusual note lengths, mid-piece key and time changes, and the key signature itself — is counted and reported in an unsupported tally. Nothing is dropped silently. It does not read another program’s proprietary project files: the route in is to export MusicXML and import that.
Export is built, and it refuses honestly until the engine attaches
The engraver accepts exactly one input — a small .ly text source read by its own parser front end — and renders it to SVG, PDF, and a Standard MIDI File. In the browser the real engine attaches only after a wasm handshake. Until it does, the export client reports itself unavailable and every export returns an explicit unavailable result: it never touches the serializer, never calls the engine, and never returns bytes. There is no code path that fabricates an export from a stub, and an engine parse or render error is contained as an error rather than a partial file.
The engraving engine underneath: real today, not a plan
Most of this page describes the suite we are still building. This section is different: it describes a piece of infrastructure that already exists and is already gate-verified -- the native engraving engine at the core of the notation side. It is owned by us, written in Rust, and carries no GPL or other copyleft code.
Its own glyph rendering, not a font subprocess
The notation symbols are vector outlines from Bravura, a SMuFL music font published by Steinberg under the SIL Open Font License, which we use under that license. On the native render path, Bravura draws every symbol it has, and the engine’s own glyph program fills only the names Bravura’s table lacks. The in-browser path, and any render where the font is absent or fails to load, draws with the owned glyph program alone. The font is read by a permissively licensed, pure-Rust font reader rather than a GPL library or an external font tool, and each symbol reaches the page as drawing paths, with no font file embedded in the PDF. No third-party font-rendering process is invoked to draw a note, a clef, or a rest.
Deterministic, integer-only engraving math
Note spacing, beam grouping, stem direction, collision avoidance, and system and page breaks are computed by the engine itself, using deterministic integer-only geometry -- the same input always produces the same output, with no floating-point drift and no hidden call-out to another notation program to do the layout for it.
Its own SVG and PDF writer
The engine writes native SVG directly, and writes PDF bytes with its own PDF writer -- not by shelling out to Ghostscript or any other external PDF-printing tool. The output format is ours end to end.
A fail-closed license gate, not a self-declared claim
Every dependency the engine draws on is checked against an explicit allowlist of permissive licenses (MIT, Apache-2.0, and similar) before the engine can ship. Copyleft, GPL-family code is not on that allowlist and cannot pass the gate. This is enforced automatically, not asserted by us after the fact.
What this is not claiming. This section describes the engine as a piece of infrastructure -- it is not a claim that the finished writing product (note entry, an editor a musician opens and works in, an in-browser app) exists today; it does not, and there is no way to buy one here. It is also not a claim about any score already published on this site or elsewhere on the platform: this engine has not been wired to produce those, they were made with separate, existing tools, and this page makes no claim that those existing scores are clean or GPL-free. The engine is a real, clean-licensed, newly built capability, and today it is not connected to that catalog.
SONUS: the multitrack audio workstation
SONUS is the audio side of the suite: where sound gets recorded, arranged, and mixed. The engine is written and the app talks to it through one narrow contract. Below is what that contract does today, and exactly where sound stops being claimed.
An owned engine behind one narrow contract
SONUS is a headless Web-Audio workstation we own, consumed by the app as an opaque module through a single narrow audio contract. The app shell never reaches into the engine: it reads one deterministic transport — position, tempo, mixer state — and that transport is the single source of truth whether or not real sound is available. Swapping the silent path for the sounding one changes no shell code.
Sound is lazy, and it is never claimed before it exists
The audio graph is imported and constructed only on the first real play, in a browser that actually has an audio context — which means after a real user gesture. With no audio context (a server render, the test runner, a headless browser) the adapter is a silent no-op transport that never throws and never imports the graph, and it reports producesSound: false. That flag flips true only once a real audio context and the graph are live. The product does not report sound it did not make.
Its own synth instruments, not a borrowed sample library
Tracks are built onto SONUS’s own subtractive synth instruments. Sampled soundfont libraries are gated separately and are not loaded by this path, and the adapter never reports a sample library it has not actually loaded. When you hear a MIDI part, you are hearing our synthesis, and the page will not pretend otherwise.
One timing conversion, written down where it happens
The score model times in ticks, where divisions is the number of ticks per quarter note. SONUS times in beats, where a beat is a quarter note. The conversion is therefore beats = ticks / divisions, and tempo is quarter-note BPM on both sides. Every delegation across the boundary converts at that one seam, so there is no second, drifting notion of time hiding in the audio layer.
MIDI in and out, honest about the hardware
Turning incoming note-on and note-off messages into notes on the roll, and turning notes back into outgoing MIDI messages, are pure functions — real, and testable with no hardware attached. The device layer is a thin wrapper over the browser’s Web MIDI entry point. Where Web MIDI is missing, or access is refused, it says so in those words — MIDI device not connected — and it never simulates a device or fabricates input to fill the silence.
The rhythm tools are real math, silent until the audio is
The step sequencer’s grid, kit selection, per-step velocity, swing offset, pattern chaining, and tempo-to-schedule timing are pure deterministic data and math — no audio, no timers, no React — so every behaviour is provable in a test. The metronome is the same: BPM to beat interval, subdivisions at the quarter, eighth, 16th, and triplet, time-signature grouping with a downbeat and configurable accents, and tap tempo averaged into a real mean rather than a guess. Playback routes each hit through the shared transport clock; while that transport reports no sound, nothing is synthesised and no click is faked. The grid, the swing, and the chain order are fully real today.
Who Musique is for
Musique is for people who both write and record music, and for the programs that teach them. A student musician can notate an assignment, hear it back, and record a performance without moving between unrelated tools. A composer or arranger can engrave a score and parts for players, then build a demo recording of the same music in SONUS. A music teacher or program can give students one suite that covers both the written and the recorded side of the craft.
The through-line is ownership of your own work. Musique is a tool you use to make music that is yours — your notes on the page, your recordings on the tracks. It is not a service that writes music for you, and nothing here is AI-generated or auto-composed.
Where the build is, honestly
Musique is not generally available today. Not generally available today That is where it stands, not a countdown: this page carries no date, no waiting list, and no sign-up. Some of what is described here already runs in code and some of it we are still building, and each section tells you which. A plan is called a plan, and no half-built piece is dressed up as a shipped, working product.
The shape of the suite is settled: a notation engraver for writing and printing music, and SONUS for recording and mixing it, under one owned roof, with the two sides meant to work together. The engraving engine underneath the notation side is real today, not a plan — an owned, GPL-free engine with its own glyph rendering, its own deterministic spacing and layout math, and its own SVG and PDF writer, behind a fail-closed license check. The engine section above walks through it. What we have not finished is the writing product around that engine: the note-entry editor a musician opens and works in, and an in-browser app. So the split is plain. The engine runs. The desk you would sit down to write at is not built.
What you can do today is write to us: tell us what you write or record, and what you need a suite like this to do. There is no price on this page, no checkout, and no account to create. A conversation is the only thing this page can start right now, so that is what it offers.
Common questions
Is Musique available to use right now?
No. Musique is not generally available today: there is nothing on this page to download, no account is provisioned from it, and no money changes hands here. What exists in code is described section by section above, each capability next to the gate that limits it. If the suite fits what you do, write to us and say so — this page carries no date and no waiting list.
What is the difference between the notation engraver and SONUS?
The notation engraver is for WRITING music: you enter notes on staves and it sets them into engraved, print-ready scores and parts. SONUS is for RECORDING music: a multitrack audio workstation where you capture audio and MIDI, edit on a timeline, and mix. One suite covers both the written page and the recorded sound.
Does Musique generate or compose music for me with AI?
No. Musique is a tool you use to write and record your own music. It does not compose, arrange, or generate music for you, and there is no AI-generated audio anywhere in the suite. Playback and virtual instruments render the notes YOU wrote or played; they do not invent new material.
Can a score I write turn into a recording?
Not in one step today. What exists now is a clean way off the notation side: its export targets are SVG, PDF, and a Standard MIDI File, so a written part leaves as something a recording session can read, instead of being stranded in one program. A direct in-suite hand-off from an engraved score to a SONUS session is a plan, not a working path.
Does it produce print-ready scores and parts?
Yes, that is what the engraver targets: clean page layout at real paper sizes, readable at a music stand, with individual instrument parts extracted from the same source as the full score. Print-ready output is the finished product the notation side is built toward.
Who is Musique built for?
People who both write and record music — student musicians, composers and arrangers, and the music teachers and programs that work with them. If you notate music AND record it, Musique is meant to keep both crafts in one place.
How much does it cost?
There is no pricing on this page and no checkout. Musique is not generally available today, so there is nothing here to buy and nothing here takes money. If you want to talk through what you need, write to us. Pricing is not something this page claims.
Is the notation engraving engine real, or is that also still a plan?
The core engraving engine is real today: an owned, GPL-free, pure-Rust engine with its own SMuFL glyph rendering, its own deterministic integer-only spacing and layout, and its own native SVG and PDF writer, gated by a fail-closed license allowlist so no copyleft code can ship in it. What does not exist today is the finished writing product around that engine -- the note-entry editor and an in-browser app -- not the engine itself.
Does this engine power the scores already published on your other pages?
No. This engine has not been wired to any score already published on this site or elsewhere on the platform. Those existing scores were produced with separate, existing tools, and this page makes no claim that they are clean or GPL-free. The engine described here is a distinct, newly built, clean-licensed piece of infrastructure, and today it is not connected to that catalog.
Why the name Musique?
Musique is the plain French word for music. It is a generic, descriptive name for a music-creation tool, chosen for exactly that reason: it says what the suite is for without dressing it up.
Related surfaces
Musique is the tool for making music. These adjacent destinations cover the wider platform and the neighbouring creative and program surfaces.
musique.center
The music building — the family home that gathers the rehearsal, writing, and library rooms in one place. Musique is the writing desk inside that building; this is the door back to the whole building.
musique.digital
The delivery side of the Musique family. This page is where you write and record; musique.digital is where every player opens their own part on the device already in their hand. Different job, same family.
musique.technology
The engraving engine underneath this suite: owned, deterministic integer-only layout with its own score writer. This page is the product a musician opens; that one is the machinery underneath it.
homeroom.software
The flagship platform brand home and the wider product family Musique belongs to. The full picture of the platform and every tool on it.
musicschool.software
Running a music program: lessons, studios, scheduling, and enrollment. Adjacent, not the same product — that runs the school; Musique is the tool a musician writes and records with.
composition.software
The writing desk for student authors — prose, revision, and the literary magazine. A sibling creative tool for text rather than music.
What this page is and is not claiming
Two things live here: a notation engraver for writing and printing scores and parts, and SONUS, a multitrack audio workstation for recording, editing, and mixing. The suite is not generally available today. No account is provisioned from this page, there is nothing here to download, and nothing here takes money. This page describes what the suite is and how much of it runs in code right now; nothing half-finished is called shipped, and no date is promised. One part is further along than the rest, and the engine section above says so on its own: the engraving engine — its own glyph code, its own deterministic layout, its own SVG and PDF writer, a fail-closed GPL-free license gate — is built and gate-verified. The writing product on top of it is not built, and the engine has not been wired to any score already published on this site. There is no AI-generated or auto-composed music in Musique. It is a tool for making your own music; playback and virtual instruments sound the notes you wrote or played, and nothing more. No competitor is named on this page. There is no price, no checkout. What you can do today is write to us.