Docs
Technical documentation
Specifications, platform compatibility and known limitations for DARKSTAR — stated plainly, kept current.
01 · Specifications
Technical specifications
Every figure below is a confirmed limit or range in the current build, not a target or an estimate.
Voices & oscillators
These are the maximums the engine supports, not a realtime guarantee. 32 voices × 3 oscillators × 16 unison lanes is a ceiling the DSP can address, but that combination does not always meet the realtime buffer deadline — actual polyphony depends on your CPU, buffer size, sample rate and patch complexity.
- Synth voices
- Up to 32
- Oscillators per voice
- 3
- Unison per oscillator
- 1–16 lanes
- Unison detune
- Up to 50 cents
- Oscillator sources
- Basic, Wavetable, Noise, Sub
- Basic waveforms
- 17
- Tuning range
- -24 to +24 semitones, ±100 cents fine
- Theoretical oscillator lane ceiling
- 1,536 lanes — a theoretical maximum, not a practical realtime target
Wavetables
- Factory wavetable
- Event Horizon — 16 frames × 2048 samples
- Custom wavetable import
- RIFF/WAVE, up to 64 frames, prepared to 2048 samples, files up to 32 MiB
Filter
- Filter types
- 8 — Low-pass, High-pass, Band-pass, Notch, All-pass, Low Shelf, High Shelf, Peaking
- Cutoff range
- 20 Hz – 18 kHz
- Resonance / Q
- 0.1–24
- Filter gain
- -30 to +30 dB
Modulation
- Envelopes
- 1 amplitude + 2 modulation (ADSR)
- LFOs
- 3, free-running, 0.01–40 Hz
- Modulation sources
- 14
- Modulation routes
- 16 per standard patch
Effects
- Effect types
- 9 — Chorus, Distortion, Reverb, EQ, Compressor, Delay, Flanger, Phaser, Metallizer
- Native DSP rack capacity
- 16 slots, serial, reorderable
- Delay time
- 1–2000 ms
- Reverb decay
- Up to 12 seconds
Factory content
- Factory patches
- 5,000
- Genre banks
- 5 — Hardcore, Techno, Industrial, Cinematic, Experimental
- Patches per bank
- 1000
- Sound-type folders
- 10 per bank (50 total)
- Factory wavetables
- 1 — Event Horizon
- Write access
- Factory presets are read-only
Output & performance
- Master gain
- -60 to +6 dB
- Glide
- 0–2000 ms, Linear / Convex / Concave
- Limiter
- None on Master — headroom is the player’s responsibility
MIDI & presets
- MIDI
- Notes, velocity, sustain, CC, channel pressure, pitch bend
- Preset format
- Versioned DARKSTAR JSON state
02 · Platforms
Platform & format compatibility
Where DARKSTAR's shared C++ DSP core currently runs, and how validated each path is.
| Form | Audio path | Status |
|---|---|---|
| Windows standalone | Native C++ via desktop audio host | |
| Browser | Same C++ core via WASM/AudioWorklet | |
| VST3 | Native C++ processor with WebView interface | |
| macOS / Linux | — |
03 · Limitations
Known limitations
Where the current build falls short of the ideal, stated up front instead of discovered the hard way.
Stellar Shift
- Noise/Sub source changes aren’t always continuous and may fall back to a non-continuous transition.
- A modulated wavetable position can block a continuous-source compile.
- FX parameter modulation across a dual-FX graph is a fallback case.
- An unresolved dual-FX morph supports a maximum of three processors per endpoint.
- Nested dual-filter or dual-FX patches must first be flattened into a plain branch.
- A saved patch restores processor configuration, but not the exact sound of an in-progress delay/reverb tail, held notes, or oscillator phase.
Performance
- The spec sheet lists 32 voices, 3 oscillators and 16 unison lanes as supported maximums — that full combination is a theoretical ceiling, not a guarantee, and can miss the realtime buffer deadline on current reference hardware.
- Practical polyphony depends on sample rate, buffer size, unison, warp, modulation and FX load.
- ARM does not yet have dedicated NEON acceleration.
- Wavetables do not yet use a pitch-dependent mip bank.
Platform & hosts
- Windows is the only platform shipping today. A macOS and Linux release is planned, but there is no confirmed date and no build yet.
- The VST3 build passes the Steinberg VST3 validator and is confirmed working in FL Studio and Reaper (Windows); acceptance across the full range of DAWs, offline bounce, the editor-closed lifecycle, multiple instances, and every sample-rate/buffer-size change is still in progress.
- The browser backend is technically functional, but broad browser compatibility and non-default output device selection are not yet fully validated.
- Native hardware MIDI and a complete standalone audio-settings page aren’t finished yet; MIDI availability in the browser currently depends partly on Web MIDI support.
Content
- The factory library ships one factory wavetable (Event Horizon); custom wavetables must be imported.
- Factory presets are read-only. Only User and explicitly added external libraries can be written to.
- There is no cloud sync, filesystem watcher, favorites system, or preset-pack installer yet.
04 · FAQ
Technical FAQ
The questions that come up once you're past setup and into how DARKSTAR actually behaves.
Yes. The engine supports up to 32 simultaneous synth voices. Heavy unison, warp, modulation and FX settings can lower practical polyphony.
Three. Each oscillator has its own source, tuning, unison, stereo position, warp and routing.
Yes. DARKSTAR imports supported RIFF/WAVE data, combines up to eight channels down to mono, and prepares up to 64 frames of 2048 samples for the engine.
Singularities focus the interface on a sound-design intent. Anomalizer creates bounded, random variations of chosen controls. Stellar Shift builds a timed transition between two complete patches — Patch A and Patch B — across up to 64 checkpoints; any checkpoint can be saved as its own patch, and the full transition can be recorded as a sound.
Not every combination. DARKSTAR analyzes five sound domains and shows which are direct, compiled, dual or fallback. Some Noise/Sub transitions and complex dual-FX combinations use a fallback path.
Yes. The rack order is the audible, serial processing order. Effects can be added, moved, bypassed and removed.
Not every DAW yet, but it is validated, not just built. The VST3 build passes the Steinberg VST3 validator and is confirmed working in FL Studio and Reaper on Windows. Acceptance across the full range of DAWs, offline bounce and every sample-rate/buffer-size change is still in progress.
There’s a working C++/WebAssembly AudioWorklet backend. Treat it as a technology preview for now.
No. The spec sheet states 32 voices, 3 oscillators and 16 unison lanes as what the engine supports as a maximum, but that full combination is a theoretical ceiling, not a promise — it does not always meet realtime deadlines. What’s actually achievable depends on your computer, buffer size, sample rate and patch complexity.
A release is planned for both, but there is no confirmed date yet and no build exists today. Windows is the only platform DARKSTAR ships on right now.
No. Master is a gain stage. High resonance, feedback, drive or a positive Master gain can clip — keep headroom.
As versioned DARKSTAR JSON patches, including synth state and, where needed, custom wavetable or morph resources. A preset doesn’t resume old held notes or an in-progress effect tail sample-for-sample.
Setting up or stuck?