# loudness
governs how perceptual loudness is distributed across time and voices. ensures dynamic balance, prevents overload, and provides a structured, deterministic way to sculpt volume at global, voice, and sectional levels.
## overview
* `0` absolute envelope
  * description: master loudness contour over time
  * real-world analogy: fade-ins, long crescendos, global swells
  * notes: defines a ceiling; all events are scaled proportionally to fit
* `1` per-voice envelope
  * description: loudness shaping per voice over time
  * real-world analogy: lead/support balance, instrument emphasis
  * notes: defines a weight profile for each voice across the timeline
* `2` region emphasis
  * description: time-local loudness emphasis based on structural divisions
  * real-world analogy: chorus/verse contrast, form-driven dynamics
  * notes: used in tandem with `segmentation` and/or `hierarchy`
## parameter behavior summary
* `0` absolute envelope
  * `a`: curvature of rise/fall (0 = linear, 1 = slow-attack sharp-release)
  * `b`: peak time (0 = early, 1 = late)
* `1` per-voice envelope
  * `a`: contrast between lead and background voices
  * `b`: voice-role skew over time (0 = fixed, 1 = rotating)
* `2` region emphasis
  * `a`: inter-region loudness contrast
  * `b`: transition smoothness between regions
all parameters are in `[0, 1]`.
## composition model
the forms are multiplicative and together produce the final per-event amplitude scaler:
```
loudness(t, v) =
    absolute_envelope(t) *
    per_voice_weight(t, v) *
    region_emphasis(t)
```
each factor is precomputed, bounded in `[0, 1]`, and strictly static - ensuring deterministic loudness for all events at all times. flat settings disable the corresponding effect (e.g. all-ones `absolute_envelope` gives full dynamic headroom).
## why these were chosen
* `absolute` enables timeline-wide energy control
* `per-voice` balances instruments or layers over time
* `region` allows macrostructure shaping through contrast and phrasing
these map directly to musical concepts (crescendo, voice lead, chorus lift) and align with a fully precomputed additive engine.
## what's not included
* runtime dynamics (compression, ducking, sidechain)
* perceptual normalization or loudness metering
* adaptive gain compensation or feedback-driven adjustment
such features would introduce non-determinism and are intentionally excluded. for expressive shifts, modulation or `mutation` domains can introduce controlled variation upstream.
## conclusion
the `loudness` domain provides compact, composable control over the perceived amplitude of sound events, while respecting a strict upper bound. it supports:
* global dynamics
* inter-voice balance
* structure-aligned phrasing
and integrates transparently with envelope and spectral shaping domains. by ensuring no overlapping events exceed the available dynamic budget, it preserves clarity, precision, and headroom at all scales.