Preset Forge

The 6 Moods

The 6D Sonic DNA

Score each dimension 0.0–1.0 based on the sound's actual character:

Engine Name vs Parameter Prefix

Engine names in "engines" and "parameters" keys use the canonical name:

Input Formats the Forge Accepts

The Forge can take any of these as input and produce .xometa output:

Format A: Guru Bin Refinement Log Table

| Change | Parameter | Old | New | Why |
|--------|-----------|-----|-----|-----|
| The Obvious Fix | filter_cutoff | 3000 | 2600 | 3kHz peak causes listening fatigue |
| The Hidden Trick | oscB_detune | 5.0 | 7.03 | Creates 1.2 Hz beating |

The Forge uses the New column for values. Ignores Old and Why for the JSON (but includes Why reasoning in the description).

Format B: Guru Bin Awakening Preset Table

## Preset: "Abyssal Memory"
Engine: Opal
Mood: Atmosphere
Parameters:
  opal_grainSize: 0.75
  opal_grainDensity: 0.6
  opal_pitch: 0.0
  ...
DNA: bright 0.2, warm 0.7, movement 0.6, dense 0.5, space 0.9, aggression 0.05

Format C: Freeform Parameter List

opal_grainSize = 0.75
opal_grainDensity = 0.6
filter at 2600
reverb mix: 0.3

The Forge interprets natural language parameter descriptions and maps them to known parameter IDs. It flags any it can't resolve and asks the user to confirm before forging.

Format D: Existing .xometa to Patch

User provides an existing preset file and a diff table. The Forge applies only the changes and writes the updated file.

The Forge Protocol

Step 1: Parse Input

Read the input — whatever format. Extract:

• Engine name (required — if ambiguous, ask)
• Parameter name:value pairs
• Any mood, DNA, description, or tag hints
• Preset name (required — if not given, propose one based on the sound character)

Step 2: Resolve Parameters

For each parameter:

1. Check if the name includes the engine prefix — if not, prepend it
2. Check that the parameter name matches a known pattern for that engine
3. Flag any unrecognized parameters with a ⚠️ and a note to verify against the engine source
4. Convert any human-readable values to normalized floats:
   • Frequency in Hz: map to 0-1 based on typical parameter range (e.g., filter cutoff 20Hz–20kHz → logarithmic scale)
   • Percentages: divide by 100
   • Already-normalized floats (0.0–1.0): pass through
   • Integer choices: pass through as integer

Step 3: Infer Missing Fields

If the user didn't specify, infer sensibly:

• Mood: from the sound character description and parameter values (high reverb → Atmosphere, heavy distortion → Flux or Aether)
• DNA: from the parameter values (high filter_cutoff → brightness 0.7+, reverb mix → space scales with reverb amount, distortion → aggression)
• Tags: 3-5 tags from the sound character (genre clues, texture words, instrument analogues)
• Description: 2-3 evocative sentences from the parameter set and any Guru Bin diagnosis text provided
• macroLabels: use the engine's canonical macro names (look these up from CLAUDE.md or ask)

Step 4: Name the Preset

Preset names follow XO_OX conventions:

• 2-3 words, max 30 characters
• Evocative, not technical (not "Low Filter Pad" — yes "Tide Before Rain")
• No jargon, no feature names
• No duplicates — read existing preset names from the destination mood folder before confirming

If the user provided a name, check it against these rules and flag any issues.

Step 5: Write the File

Output the .xometa file to:

~/Documents/GitHub/XO_OX-XOmnibus/Presets/XOmnibus/{Mood}/{PresetName}.xometa

Or if it's an engine-specific preset (not XOmnibus-level):

~/Documents/GitHub/XO_OX-XOmnibus/Source/Engines/{EngineName}/Presets/{Mood}/{PresetName}.xometa

Always ask the user to confirm the destination path before writing.

Step 6: Validate

After writing, run a quick validation:

• JSON parses without error
• schema_version is present
• engines array is non-empty and uses canonical names
• parameters keys match engine names in engines array
• DNA values are all 0.0–1.0
• No parameter values obviously out of range (negative when they shouldn't be, > 1.0 when normalized)
• Preset name is unique in the destination folder

Report the validation result. Flag any issues.

Batch Mode

When Guru Bin produces a full Awakening Preset set (5-10 presets from a Retreat), the Forge processes all of them in sequence:

1. Read all preset descriptions from the retreat document
2. Parse each one
3. Infer all missing fields
4. Propose all names as a batch for user review
5. User confirms or revises
6. Write all files
7. Report: files written, paths, any flags

Coupling Presets

When a preset involves two engines (coupling preset in Presets/XOmnibus/Entangled/):

• "engines" array contains both engines: ["Opal", "Overdub"]
• "parameters" has keys for both engines
• "coupling" section describes the pairs:

"coupling": {
  "pairs": [
    {
      "source": "Opal",
      "target": "Overdub",
      "type": "FilterFrequencyModulation",
      "amount": 0.25
    }
  ]
}

Available coupling types (from MegaCouplingMatrix.h): FilterFrequencyModulation, AmplitudeModulation, PitchModulation, WaveshapeModulation, TemporalModulation, SpectralShaping, EnvelopeFollowing, GrainTrigger, ResonanceControl, PhaseModulation, AudioToBuffer, RingModulation

Alternative Output: C++ setParam() Code

If the user needs C++ code instead of (or in addition to) a .xometa file — for example, to hardcode a factory preset in an engine's PluginProcessor — the Forge can generate:

// Preset: "Abyssal Memory" — Opal — Atmosphere
void loadPreset_AbyssalMemory(juce::AudioProcessorValueTreeState& apvts)
{
    auto setParam = [&](const juce::String& id, float value) {
        if (auto* p = apvts.getParameter(id))
            p->setValueNotifyingHost(value);
    };

    setParam("opal_grainSize",    0.75f);
    setParam("opal_grainDensity", 0.60f);
    setParam("opal_pitch",        0.00f);
    // ...
}

Invoke this with: "Forge as C++" or mode: cpp.

Arguments

• (none) — the Forge reads the current conversation for parameter input
• mode: json (default) — output .xometa file(s)
• mode: cpp — output C++ setParam() function(s)
• mode: both — output both formats
• batch — process multiple presets from a retreat document
• patch — modify an existing .xometa file rather than creating new