Summaries
The charts, in words. Every Fingerprint (and every comparison) can carry a short plain-language description of what the numbers say: how it clips, whether it cleans up, what chords will do. The summary describes the whole acquisition, so it renders in exactly two places — the Fingerprint’s overview page and the pedal overview; the individual chart pages link to it rather than repeating it. This page explains where the words come from — and, just as important, where they are not allowed to come from.
Where the words come from
Every sentence in a summary is generated from numbers the measurement pipeline computed — a knee level, an even-to-odd harmonic balance, a cleanup point. The app first extracts those features, then turns them into prose with fixed, hand-written templates (“generated from measurements”): deterministic wording, no generative model anywhere in the path, so the same measurement always produces the same words. A summary can never say anything the numbers don’t support — the Evidence disclosure under each summary lists exactly the values the words were built from.
A null run — a control capture of the loop with no device in it — gets a summary in a different register entirely. Its subject is the rig, so the words are about the rig: what the loop itself contributes at the measured drive, and what that means for readings taken through it (“distortion readings near this level are the rig, not a device”). The pedal-voice sections — clipping, touch dynamics, chords — are simply absent, because a straight wire has none of those things to describe, and the brightness/texture proxies are never computed for it at all. Tone-review prose about a cable would read as parody, and a summary that can be laughed at on a wire has not earned trust on a pedal.
Three layers, and each is honest about its job:
Features — deterministic numbers from the measurement products (harmonic distortion, compression, transfer curve, chord IMD, gain map). Whatever wasn’t measured simply isn’t there: a record with no two-tone measurement gets a summary that says nothing about chords.
Perceptual metrics — brightness (“sharpness”) and texture (“roughness”) computed on the same loudness-matched renders the Hear-it panels play. These are simplified proxy implementations of published psychoacoustic models — good for comparisons, not certified absolute values, which is why their evidence rows say “(proxy)”.
Wording — a versioned lexicon maps feature ranges to the vocabulary the description may use, and deterministic templates assemble the sentences. No generative model is involved anywhere: the wording layer verbalizes the evidence, it never analyzes or invents. The chip under every summary says so — Generated from measurements.
Settings (⌘,) → Visualization holds the Show summaries toggle (hide them entirely if you’d rather keep the charts to themselves) and, beneath it, the grounding contract in one line: every number in a summary must trace to a measurement shown in its Evidence disclosure.
Because the wording is deterministic, a summary only ever needs computing once: the extracted features are cached (keyed on the measurements’ actual content, so any change to the underlying data produces a fresh summary automatically), generation starts in the background as soon as a measurement is saved, and revisiting a page shows the summary instantly instead of re-deriving it. The Regenerate button is the explicit exception — it discards the cached features and re-extracts everything from the stored measurements, which on identical data produces identical words (that determinism is the point).
For the curious: the validator
The project’s test suite holds the wording layer to its own rules with plain code: every template’s prose is checked against its own evidence packet — each number in the text must appear there (rounding is tolerated; fabrication is not), and circuit-topology claims — component names, chip types, clipping-element chemistry — are rejected outright, because a black-box measurement cannot license them. Deliberately corrupted prose is part of the suite, so the validator itself is tested.
Reading a comparison
Comparison summaries are built from both records’ evidence plus their computed differences, so their claims are relational — “clips about nine dB later”, “carries more even-harmonic content” — which restates the measurements directly. The last line, After loudness matching, is the one to trust most: it lists only the differences that should survive once both devices are matched to equal loudness, because louder-sounds-better is the oldest trick in gear comparison.
What the caveats mean
Summaries hedge on purpose. A feature extracted from a short live-mode sweep, a hysteresis reading with no same-setting Harmonic Distortion measurement to phase-compensate it, or a proxy metric all carry reduced confidence — the wording softens (“appears to”, “may”) and a caveat line says why. A caveat is not the app being coy; it is the measurement telling you how far to trust the words.
The words are heuristics
The mapping from numbers to vocabulary is editorial judgment informed by circuit theory and psychoacoustics — not listener-validated ground truth. Nobody has run a listening panel to confirm that this knee sharpness reads as “abrupt” to most players. The lexicon is versioned (every summary records the version that worded it) so the mapping can be calibrated against real listening data someday without silently changing old descriptions’ meaning.
Common misreadings
A summary is not a review. “True limiter” and “clean bleed” are measured behaviors, not verdicts; which one you want is a musical choice.
Absent prose is not a shrug. If the chord section is missing, the measurement to support it doesn’t exist yet — complete the fingerprint’s Chord IMD sub-measurement and the sentence appears.