SoundScript is an independent open-source audio and media programming language. Write timed visuals, melodies, phrases, orchestration, vocal cues, and industrial sonification as plain text; compile them to deterministic MIDI, WAV, or a browser-playable audio-visual clip with the .NET CLI or a client-side browser playground.
block intro { phrase { mf play arp Cmaj q } } pattern arp { up } track melody { layer piano gain 0.9 play intro }
Hear and watch repository-generated outputs. Explore the audio sources below, or play and edit the Audio/Visual demo in the Playground.
A piano-backed timeline with sequential cues, an expanding orb, overlapping sparkle, and an outro fade.
sync audio
visual "intro" for 3s
visual "circle" for 8s at 0s {
animate radius 28 -> 170 over 3s
}
The full source includes the piano track, all cues, and animations.
One piano track with phrase shaping and legato articulation.
Inputtempo 112 instrument piano C4 q E4 q G4 q C5 q
tempo 112
time 4/4
instrument piano
track piano {
phrase {
curve gentle
articulation legato
mf
C4 q E4 q G4 q C5 q
G4 q E4 q D4 q C4 q
}
}
Violin, cello, and piano rendered from three deterministic tracks.
Inputtrack strings { instrument violin … }
track harmony { instrument cello … }
track pulse { instrument piano … }
tempo 92
time 4/4
track strings {
instrument violin
mp
C4 q E4 q G4 q C5 q
A4 q F4 q D4 q G4 q
}
track harmony {
instrument cello
p
C3 h G3 h
F3 h G3 h
}
track pulse {
instrument piano
mp
Cmaj h Fmaj h
Amin h Gmaj h
}
V9 Jingle Bells mix rendered from the bundled WordBank human-audio corpus and deterministic G2P pipeline.
Inputspeak "Jingle bells jingle bells" speak "Jingle all the way" speak "Oh what fun it is"
tempo 132
time 4/4
track melody {
instrument violin
mf
phrase { curve soft transition smooth play hook }
}
track harmony { instrument piano p Cmaj w Cmaj w }
track bass { instrument bass mf C2 w C2 w }
speak "Jingle bells jingle bells" seed=7 gain=1.05
speak "Jingle all the way" seed=8 gain=1.05
speak "Oh what fun it is" seed=9 gain=1.0
The engineering surface is deliberately small: plain text in, standard files out, with repeatability enforced by tests.
Version timed cues and property curves alongside a musical score. Scrub exact states, preview synchronized audio, and export a WebM clip. Timeline states and source PCM are deterministic; encoded video bytes can vary by encoder.
Identical scripts produce byte-identical MIDI. Seeded humanization and audio regression checks preserve reproducibility across runs and platforms.
The CLI, MIDI compiler, Wave renderer, timbre engine, and WordBank pipeline run locally. The browser playground runs client-side with no account.
Compositions are reviewable plain-text .ss and .ssw files that can be diffed, versioned, and rebuilt like code.
Compile to MIDI with karaoke lyric events, render direct WAV, or use SoundCSS for deterministic MIDI-to-WAV/OGG timbre synthesis.
| Property | Traditional DAW | AI Generator | SoundScript |
|---|---|---|---|
| Reproducibility | Depends on project, plug-ins, and versions | Depends on model, seed, and service version | Byte-stable MIDI/PCM and deterministic visual states; encoded WebM varies by encoder |
| Git friendly | Project-format dependent | Prompts can be versioned; outputs are media files | Plain-text scripts are diffable |
| Offline | Common after installation | Product and model dependent | CLI and synthesis pipelines run locally |
| Deterministic | Depends on plug-ins and workflow | Model and seed dependent | Explicit product invariant |
| Programmable | Automation and scripting vary | API availability varies | Purpose-built language and C# libraries |
| Automation | Host-specific | Service-specific | Cross-platform CLI and text files |
| Media / Audio-Visual | Video and synchronization workflows vary by host | Audio/video generation and editing vary by model | Text-authored intervals, overlays, linear property curves, shared audio clock, and browser or CLI WebM export |
| Accessibility | Interface and host dependent | Service interface dependent | Text input, CLI, and keyboard-operable browser UI |
The repository includes runnable .ss and .ssw examples spanning composition, vocals, Wave synthesis, and industrial cues.
visual "intro" for 3s
visual "circle" for 8s at 0s {
animate radius 28 -> 170 over 3s
}Author deterministic state over time, then sample it in the Playground at 24, 30, or 60 FPS to export a WebM clip with the shared SoundScript.Wave PCM audio rail.
Visual programs remain a deterministic StateAt(t) function. The Playground samples that function only in its WebM renderer, combines it with the shared SoundScript.Wave PCM audio rail, and downloads a playable clip.
tempo 120 → 140 over 2 bars
block verse { phrase { curve balanced play arp Cmaj q } }
track melody { layer flute layer cello play verse }
voice lead {
vocal choir
sing "Twinkle twinkle little star"
C4 q C4 q G4 q G4 q
}
block idle { phrase { curve gentle articulation legato … } }
block critical { phrase { transition sharp articulation staccato … } }
track machine { play idle · play running · play critical }
track spatial {
phrase { curve gentle articulation legato … }
phrase { transition sharp articulation staccato … }
}
phrase { curve swell transition expressive mf C4 q E4 q G4 q }
phrase { curve fade decrescendo f C5 q G4 q E4 q }
pattern arp { up }
pattern strumPat { strum }
track melody { play arp Cmaj q play strumPat Cmaj q }
Timed title cards, motion studies, and layered product cues with a scripted score. Explore use cases.
Versioned compositions, CLI workflows, and reusable C# libraries.
Repeatable rule-based audio experiments and inspectable pipelines.
Plain-text notation, harmony, orchestration, and synthesis examples.
Text and keyboard workflows plus non-speech spatial cue examples.
Versionable machine-state, timing-drift, and process-trend cues.
MIDI, Wave, SoundCSS, phonetics, and programmatic arrangement.
Deterministic cue generation for repeatable interactive behavior.
V10 adds continuous-time visual programs alongside the existing audio language. V11 projects those states into a shared scene for live playback and browser or CLI WebM export.
StateAt(t).The current presentation profile renders an intro pill, circle/orb, product card, sparkle/star, and generic named cards. These are built-in treatments, not imported image or video assets.
Try the Audio/Visual library → · Explore 20 practical compositions · Read the supported constructsThe phrase engine maps machine data to non-speech cues using curves, articulation, dynamic envelopes, and timing modifiers. Cue libraries remain plain text and can live in version control beside system code.
SoundScript is an MIT-licensed engineering platform for writing audio like code. Notes, chords, dynamics, patterns, orchestration, vocals, and industrial cues pass through explicit, inspectable stages rather than a generative model.
It runs on Windows, macOS, and Linux through .NET 10, plus WebAssembly in Chrome, Edge, Firefox, and Safari. No server, account, DAW, or plug-in is required for the core workflow.
SoundScript is created by Anup Jayant Dharangutti, a music enthusiast and architect of enterprise-grade automation systems. He also created UniverseGlance and AutomationGlance.
A shared parser builds the AST. Independent output rails then produce MIDI, direct Wave audio, vocal-enhanced mixes, or temporal visuals with synchronized media export.
Exact time intervals, absolute overlays, linear automation, audio synchronization markers, and a tempo-aware beat bridge. Play, pause, resume, restart, and scrub the same compiled timeline.
One canonical scene profile and fitted Wave PCM track feed live preview, browser WebM, and CLI WebM. Output FPS belongs to the export adapter. CLI encoding requires FFmpeg; browser export requires canvas capture and MediaRecorder.
Offline, deterministic human-audio vocalization combines a curated CC0/CC-BY corpus with rule-based G2P timbre. V9 adds the Jingle Bells WordBank preset and corpus coverage in the Playground workflow.
Render .ss or .ssw directly to deterministic WAV. Mix recordings through sample / speak sample=, or generate stems with soundscript vocal and wave --offline-tts.
SoundCSS controls deterministic MIDI-to-WAV/OGG synthesis. Per-cycle harmonics, formant Q and drift, band-shaped noise, transient modeling, and crossfading replace flat frame snapshots.
PhonemeComposer maps text to syllables, phonemes, gestures, and MIDI. ProsodyComposer plans phrase, word, and syllable pitch. --emit-ss round-trips composed output to editable source.
sync audio
visual "intro" for 2s
wait 500ms
visual "circle" for 3s at 0s {
animate radius 24 -> 120 over 3s
}Time-based source; V11 adds downstream WebM rendering. All media constructs and properties
import "lib.ss"
block intro { ... }
play intropattern arp { up }
play arp Cmaj q
phrase { curve soft }gain 0.9 humanize 0.03 layer piano layer cello
double octave Cmaj drop2 tempo 120 → 140 over 4 bars
voice lead {
vocal choir
sing "Twinkle twinkle" C4 q C4 q G4 q G4 q
}Lyrics bind to notes deterministically and export as karaoke events. Vocal docs
compose "Twinkle twinkle little star"
Identical text yields identical MIDI bytes through the rule-based PhonemeComposer. Composer reference
git clone https://github.com/dharangutti/sound-script.gitcd sound-script && dotnet builddotnet run --project src/SoundScript.Cli -- run examples/full-v2-showcase.ssdotnet run --project src/SoundScript.Cli -- compose "Twinkle twinkle little star"dotnet run --project src/SoundScript.Cli -- compose "Twinkle twinkle little star" --emit-ss twinkle.ssdotnet run --project src/SoundScript.Cli -- run examples/vocal-song.ssdotnet run --project src/SoundScript.Cli -- wave examples/jingle-bells-wordbank.ssw jingle.wav --tts-dir vocal-stems/wordbankLanguage reference
CLI reference
Notation engine
Expressive notation
V11 media export
V10 timing and supported primitives
Audio/Visual use cases
Temporal media export
Audio/Visual Playground
Runnable examples
Architecture
Interpreter pipeline
Timbre engine
v1.2 foundation
sample / speak sample=, or generate offline stems with soundscript vocal and wave --offline-tts..ss / .ssw directly to deterministic WAV through speak, effects, and seeded humanization.--emit-ss serializes composed output to editable source that round-trips to byte-identical MIDI.prosody CLI verb.compose verb turn plain text into melody through syllables, phonemes, and musical gestures.