ENGINE=voice · src/engine/voice/ · generated from osc.dsp + filter.dsp + voice.json

A drone voice - and the worked example of the series (chain) dual-deck generated path: two different Faust kernels wired in sequence, with deck A driving the first stage and deck B the second. Here stage 1 is a drone oscillator (an instrument - it generates sound from its knobs with no audio input) and stage 2 is a resonant filter, so the signal flows osc → filter → out. Authored as two .dsp files plus a JSON manifest with no hand-written C++ (the generator emits the wrapper on the shared FaustChainEngine<Traits> template). See docs/dev/engine-gen.md §9 for the dual-deck design and docs/engine-types/faust.md for the Faust path.

Controls

_Generated from docs/diagrams/controls/voice.json via make diagrams._

Deck A drives the oscillator (stage 1):

Deck B drives the filter (stage 2):

The same physical knobs mean different things on each deck because each deck drives its own stage. The chain is mono internally and duplicated to the stereo bus; the LED rings show an output-level meter. As an instrument, it makes sound on its own - no input patch needed.

Authoring (the series dual-deck path)

Three files in src/engine/voice/:

• osc.dsp - stage 1, a 0-input drone oscillator (freq/shape/level).

• filter.dsp - stage 2, a resonant low-pass (cutoff/reso/drive/mix).

• voice.json - the manifest. "deck_mode": "series" lists the two stages (stage 1 → deck A, stage 2 → deck B), each with its own knob map:

{ "engine": "voice", "backend": "faust", "deck_mode": "series",
  "stages": [
    { "dsp": "osc",    "knobs": { "Pitch": "freq", "Size": "shape", "Mix (SOS)": "level" } },
    { "dsp": "filter", "knobs": { "Pitch": "cutoff", "Position": "reso", "Size": "drive", "Mix (SOS)": "mix" } }
  ],
  "features": { "meter": true, "color": "0x88ff33" } }

Then:

make faust-engine MANIFEST=src/engine/voice/voice.json   # both kernels + wrapper + build + diagram
make -j8 ENGINE=voice                                  # build (~84% SRAM_EXEC)
make engine-voice                                      # clean + build + DFU flash
make -C host test-voice                                # host test (chain + per-stage independence)

Stage A may have inputs (an FX→FX chain) or none (instrument→FX, as here). The two stage kernels get engine-scoped namespaces (fx_voice_osc, fx_voice_filter) so stage names can't collide across engines. voice_engine.h and the generated kernels are checked in; regenerate the wrapper with --force-glue.