spec: "1.0"
id: "com.grooveforge.autotune"
name: "Autotune"
version: "0.1.0"
type: effect

# ─── Parameters ────────────────────────────────────────────────────────────
# Physical units throughout; the host converts from the normalised knob
# position before sending them to the native DSP (gfpa_dsp.cpp,
# AutotuneEffect).
#
# Defaults are the full-on robot: every note, instantly, all the way. That is
# the sound people put an autotune on for; the knobs are there to back it off.
parameters:
  # The key and scale notes are pulled towards. A scale patched into the
  # SCALE IN jack (Xen, Jam Mode) overrides both. The scale itself arrives
  # from the host as `scale_mask`, which is deliberately not a parameter: it
  # is not the user's setting and must not be saved with the project.
  - id: key
    paramId: 0
    name: "Key"
    min: 0
    max: 11
    default: 0
    type: selector
    options: ["C", "C#", "D", "D#", "E", "F", "F#", "G", "G#", "A", "A#", "B"]

  # Order matches gf_autotune_scale in native_audio/gf_autotune.h.
  - id: scale
    paramId: 1
    name: "Scale"
    min: 0
    max: 6
    default: 0
    type: selector
    options: ["Chromatic", "Major", "Minor", "Harmonic Minor", "Major Penta", "Minor Penta", "Blues"]

  # How much of the distance to the target note is corrected.
  - id: strength
    paramId: 2
    name: "Strength"
    min: 0
    max: 100
    default: 100
    unit: "%"
    display: value

  # How fast the correction reaches the note. 0 ms is the hard, stepped,
  # robotic sound; 50 ms and up starts to sound like a singer who is simply
  # in tune.
  - id: retune
    paramId: 3
    name: "Retune"
    min: 0
    max: 400
    default: 0
    unit: "ms"
    display: value

  # Slows the correction on held notes only, so their vibrato survives while
  # short notes still snap.
  - id: humanize
    paramId: 4
    name: "Humanize"
    min: 0
    max: 100
    default: 0
    unit: "%"
    display: value

  # Only corrects notes that are already near a scale note. Slides, scoops
  # and deliberate blue notes further away pass through untouched.
  - id: flex_tune
    paramId: 5
    name: "Flex-Tune"
    min: 0
    max: 100
    default: 0
    unit: "%"
    display: value

  # Fixed shift after correction. An octave down is the monster, an octave up
  # the chipmunk.
  - id: transpose
    paramId: 6
    name: "Transpose"
    min: -12
    max: 12
    default: 0
    unit: "st"
    display: interval

  - id: mix
    paramId: 7
    name: "Mix"
    min: 0
    max: 100
    default: 100
    unit: "%"
    display: value

# ─── Signal graph ──────────────────────────────────────────────────────────
# Pass-through. The correction runs natively (AutotuneEffect in gfpa_dsp.cpp)
# on every platform that has a native audio engine; this graph only exists so
# the descriptor validates and a host without one leaves the signal untouched.
graph:
  nodes:
    - id: in
      type: audio_in
    - id: out
      type: audio_out
  connections:
    - from: in.out
      to: out.in

# ─── UI ────────────────────────────────────────────────────────────────────
ui:
  layout: lanes
  groups:
    - label: "Notes"
      controls:
        - type: selector
          param: key
          label: "Key"
        - type: selector
          param: scale
          label: "Scale"

    - label: "Tune"
      controls:
        - type: knob
          param: strength
          label: "Strength"
          size: small
        - type: knob
          param: retune
          label: "Retune"
          size: small
        - type: knob
          param: humanize
          label: "Humanize"
          size: small
        - type: knob
          param: flex_tune
          label: "Flex"
          size: small

    - label: "Voice"
      controls:
        - type: interval
          param: transpose
        - type: knob
          param: mix
          label: "Mix"
          size: small