feat: one knob for the five effects that cannot honestly have one (#3188)

* fix(studio): close the typecheck and fallow gaps wa-18b-reschedule opened

useAutomationLanes.ts's write() assumed gesture-scoped coalescing and a
preview-only commit that useDomEditAttributeCommits.ts never grew — backported
that option support from its own later commit so the two sides of the API
agree. The paste path and its tests were missing the box selection's v0/v1
bounds a sibling commit added to AutomationSelection. The FX panel's carve
controls still edited the six mechanism numbers (maxCutDb, bands,
intelligibilityBias) after carveProfile() collapsed authoring to one Strength
knob, so those fields no longer existed on HfCarveSettings; UI now edits
strength, and analyseCarveBands is called with carveProfile(strength).

Also closes fallow's complexity, dead-code and duplication findings on this
PR's diff: extracted automationLaneDragMath.ts (pure group/point-move math)
and useAutomationRangeDrag.ts (the marquee-select gesture) out of
useAutomationLaneGestures.ts, pulled a couple of render-loop ternaries and a
resolver into named functions, dropped an export nothing outside its file
used, and shared a step-simplifier between audioCarve's two envelope
builders.

The edge-stretch vs. box-select priority test in TimelineAutomationLane.test
was still pinning the pre-box-select rule (edge wins over a point sitting on
it) that a sibling commit deliberately reversed — a point inside the box is
now selected content, so grabbing it drags the group instead. Updated the
test to the shipped rule instead of the old one.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* fix(core): cap the via conic's weight so an edge-clamped via point can't NaN

A via point pulled out past the segment (viaX: 5, viaY: -3) clamps to
(0.999, 0.001) — exactly on the steady region's edge, where edge - viaX is 0.
viaConic divided by that zero to get an infinite weight, and shapeVia turned
Infinity into NaN a few steps later (Infinity - Infinity in the quadratic
coefficient). NaN reaching setValueCurveAtTime silences the automated
parameter for the rest of the render.

Capped the weight at 1e6 instead of leaving it unbounded — past that point
the arc already reads as touching the via point, so nothing visible is lost.
Also hardened shapeVia's existing denominator guard (`<= 0`) to `!(> 0)`,
since NaN fails the original comparison and fell through it.

Review by Miga (PR #3208).

* fix(studio-server): fingerprint the proactive waveform cache key too

The route already keys the waveform cache on the asset's size and mtime as
well as its path, so a rebuilt-in-place file gets fresh peaks instead of
stale ones. generateWaveformCache — the proactive path that runs on upload —
still called buildWaveformCacheKey with the path alone, so it wrote to a
different key than the route reads from (making the pre-generated cache
never found) and kept the exact collision bug this fingerprint exists to fix
on its own path.

Review by Miga (PR #3211).

* style(docs): run oxfmt on the /hyperframes-audio skill docs

Table column widths had drifted out of alignment with oxfmt's own rules,
failing format:check and blocking the Preflight gate every downstream
branch inherits. Whitespace only, no content change.

* fix(core): stop \b from missing underscore-separated names, guard clipsOverlap's negative duration

\b treats `_` as a word character, so \bbed\b never matched bed_01,
music_bed_loop, or theme_song, and \bvo\b/\bvox\b/\btts\b had the same gap —
an underscore-separated bed classified as "unknown" and could end up offered
as its own carve source. Replaced the short hints with a boundary that
actually excludes letters and digits on both sides.

clipsOverlap computed end = start + duration without guarding sign, so a
negative duration put end before start — an interval that does not describe
anything, and one specific case showed it silently dropping a real overlap
(a shorter, earlier broken end rejected a clip that genuinely contained the
point). Duration clamps to zero instead: a clip cannot un-play time, and a
zero-length clip at its start is the sane reading of "duration nobody wrote
down as positive."

Review by Miga (PR #3212).

* fix(studio): widen PropertyPanel's resetModules render timeout again

The 20s margin (already once widened for the same reason) is timing out in
CI's full-monorepo Test run — the resetModules()+fresh-import render this
test needs is uncached and competes with every other package's test suite
for the same worker pool, and the same test passes in well under 2s
standalone. Went to 45s rather than re-tuning to whatever number happens to
clear the current CI load, since that number moves every time CI gains a
package.

* fix(studio): stop the single-candidate auto-apply carve firing twice

Two auto-apply effects both fire when sourceOptions.length === 1: the
multi-candidate effect only guards length === 0, so a single candidate
passes it too, and the single-candidate effect passes its own guard right
after — both compute the same sources list and both call setCarve, so the
common case (one narrator, one bed) triggered two decodes, two FFT runs, and
two concurrent attribute writes for one decision.

The multi-candidate effect now defers to its sibling for exactly one
candidate, which already has its own detailed handling for that case.

Review by Miga (PR #3213).

* feat(core): carve against every voice over a bed, always (#3212)

* feat(core): carve against every voice over a bed, always dynamically

A bed usually runs under a whole sequence — a narrator, an interview answer, a
second presenter — and carving against one of them left the others fighting it.
`source` becomes `sources`, and `mixCarveSources` sums every voice onto the BED's
clock before anything is measured. That is what keeps one analysis sufficient: the
chain is fixed, so there is no per-voice filter to switch between, and bands drawn
from all the speech there is with envelopes that rise wherever any of it happens
answer the actual question — where and when is speech masking this bed.

Summed rather than averaged: two people talking at once mask more than either
alone. Audio before the bed starts is dropped rather than folded in at zero, since
it plays over nothing and shifting it would put a cut where there is no voice.

`dynamic` is gone. A fixed depth thins the bed through every pause, and once both
have been heard there is no reason to want it, so every carve follows the speech.

Two helpers the panel and the headless script now share instead of each carrying a
copy — two definitions of "what does this name suggest" drift, and then the two
disagree about which track is the voice:

- `classifyAudioName` reads a track's kind from its id and filename together.
  `unknown` is deliberately common: treating an unrecognised name as "not a voice"
  would hide the one track somebody needs to pick.
- `clipsOverlap` keeps out a voice that never plays while the bed does. An unwritten
  duration counts as unbounded, not zero — refusing a clip whose length the
  composition leaves to the media would drop the commonest case there is.

Files written before this still load: a single `source` reads as a one-voice list,
a stored `dynamic` is ignored, and an absent attribute means the defaults whole.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* fix(core): stop \b from missing underscore-separated names, guard clipsOverlap's negative duration

\b treats `_` as a word character, so \bbed\b never matched bed_01,
music_bed_loop, or theme_song, and \bvo\b/\bvox\b/\btts\b had the same gap —
an underscore-separated bed classified as "unknown" and could end up offered
as its own carve source. Replaced the short hints with a boundary that
actually excludes letters and digits on both sides.

clipsOverlap computed end = start + duration without guarding sign, so a
negative duration put end before start — an interval that does not describe
anything, and one specific case showed it silently dropping a real overlap
(a shorter, earlier broken end rejected a clip that genuinely contained the
point). Duration clamps to zero instead: a clip cannot un-play time, and a
zero-length clip at its start is the sane reading of "duration nobody wrote
down as positive."

Review by Miga (PR #3212).

---------

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

* fix(studio): port the carve UI off the removed source/dynamic fields

#3212 (accidentally squash-merged into this branch instead of main) changed
HfCarveSettings from a single `source` + `dynamic` toggle to a `sources`
list with dynamic mode removed outright — the multi-voice UI consumer that
goes with that shape lands in the very next PR, so this branch was left with
a type that no longer matched its own code.

Minimal port, not the multi-voice redesign that PR does properly: the
"Listen to" picker and analyse() treat sources[0] as the one voice this UI
still understands, and every dynamic-mode branch (the automated envelope
lanes, the toggle, the checkbox) is gone along with the field — a carve is
now always the static value the analysis computes, matching what the type
change made permanent. Test suite trimmed the same way: the automation-lane
and toggle tests covered behavior that no longer exists.

---------

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

packages/core/src/audioFxProfiles.test.ts

@@ -0,0 +1,122 @@
+import { describe, expect, it } from "vitest";
+import { defaultAudioFxParams, getAudioFxDef } from "./audioFx.js";
+import { EFFECT_COPY } from "./audioFxCopy.js";
+import {
+  applyAudioFxProfile,
+  audioFxProfileStrength,
+  getAudioFxProfile,
+  HF_AUDIO_FX_PROFILES,
+} from "./audioFxProfiles.js";
+
+describe("derived one-knob profiles", () => {
+  it("covers exactly the effects whose copy asks for one", () => {
+    // `primary: "strength"` is the copy layer saying "this module has no real
+    // parameter that can honestly be its face". A profile missing for one of
+    // those leaves the module opening on all seven of its controls; a profile
+    // for anything else is a knob nobody asked for.
+    const asked = Object.entries(EFFECT_COPY)
+      .filter(([, copy]) => copy.primary === "strength")
+      .map(([id]) => id)
+      .sort();
+    expect(Object.keys(HF_AUDIO_FX_PROFILES).sort()).toEqual(asked);
+  });
+
+  it("derives only parameters the effect actually has", () => {
+    for (const [id, profile] of Object.entries(HF_AUDIO_FX_PROFILES)) {
+      const keys = getAudioFxDef(id)?.params.map((p) => p.key) ?? [];
+      for (const key of profile.derives) {
+        expect(keys, `${id} derives "${key}", which it does not have`).toContain(key);
+      }
+      // And the curve sets everything it claims to.
+      for (const key of profile.derives) {
+        expect(profile.at(0.5)[key], `${id}.${key} is not set at 0.5`).toBeDefined();
+      }
+    }
+  });
+
+  it("stays inside every parameter's declared range across the whole knob", () => {
+    // A profile that runs past a range is not clipped by the panel — it is
+    // clamped on the way into the chain, so the knob would silently stop
+    // meaning anything past that point.
+    for (const [id, profile] of Object.entries(HF_AUDIO_FX_PROFILES)) {
+      const def = getAudioFxDef(id);
+      for (let s = 0; s <= 1.0001; s += 0.05) {
+        const derived = profile.at(s);
+        for (const param of def?.params ?? []) {
+          const value = derived[param.key];
+          if (value === undefined || param.kind !== "number") continue;
+          expect(value, `${id}.${param.key} at ${s.toFixed(2)}`).toBeGreaterThanOrEqual(param.min);
+          expect(value, `${id}.${param.key} at ${s.toFixed(2)}`).toBeLessThanOrEqual(param.max);
+        }
+      }
+    }
+  });
+
+  it("moves every derived parameter monotonically", () => {
+    // The whole promise of one knob: everything under it moves together and in
+    // one direction. A parameter that turns around mid-sweep means the knob
+    // makes the effect stronger and then weaker, which is unusable and is the
+    // failure a hand-tuned table invites.
+    for (const [id, profile] of Object.entries(HF_AUDIO_FX_PROFILES)) {
+      for (const key of profile.derives) {
+        const series: number[] = [];
+        for (let s = 0; s <= 1.0001; s += 0.05) {
+          const value = profile.at(s)[key];
+          if (typeof value === "number") series.push(value);
+        }
+        const up = series.every((v, i) => i === 0 || v >= (series[i - 1] ?? v));
+        const down = series.every((v, i) => i === 0 || v <= (series[i - 1] ?? v));
+        expect(up || down, `${id}.${key} turns around mid-sweep`).toBe(true);
+      }
+    }
+  });
+
+  it("keeps what the author set under Details", () => {
+    // A profile names only the parameters it derives. A compressor's knee and a
+    // saturation's curve type are the author's, and the knob must not reach in
+    // and reset them on the way past.
+    const params = { ...defaultAudioFxParams("compressor"), knee: 6, mix: 0.5 };
+    const next = applyAudioFxProfile("compressor", 0.8, params);
+    expect(next.knee).toBe(6);
+    expect(next.mix).toBe(0.5);
+    expect(next.ratio).not.toBe(params.ratio);
+  });
+
+  it("reads a strength back out of the values a chain stores", () => {
+    // A chain stores mechanism, not the knob — the mechanism is what renders,
+    // so it has to be authoritative. Reopening a project has to put the knob
+    // back where it was.
+    for (const [id, profile] of Object.entries(HF_AUDIO_FX_PROFILES)) {
+      for (const s of [0, 0.25, 0.5, 0.75, 1]) {
+        const params = applyAudioFxProfile(id, s, defaultAudioFxParams(id));
+        expect(audioFxProfileStrength(id, params), `${id} at ${s}`).toBeCloseTo(s, 1);
+      }
+      void profile;
+    }
+  });
+
+  it("passes through the figures the design proposed", () => {
+    // The curves replaced a three-point table, and these are the points it
+    // named. Continuous beats three settings, but not at the price of landing
+    // somewhere else than the design said.
+    expect(getAudioFxProfile("compressor")?.at(0).threshold).toBe(-12);
+    expect(getAudioFxProfile("compressor")?.at(0.5).ratio).toBe(4);
+    expect(getAudioFxProfile("compressor")?.at(1).release).toBe(90);
+    expect(getAudioFxProfile("gate")?.at(0).threshold).toBe(-55);
+    expect(getAudioFxProfile("saturate")?.at(1).threshold).toBe(-18);
+    expect(getAudioFxProfile("reverb")?.at(0).size).toBe(0.25);
+    expect(getAudioFxProfile("reverb")?.at(0.5).size).toBe(0.55);
+    expect(getAudioFxProfile("reverb")?.at(1).size).toBe(0.9);
+    expect(getAudioFxProfile("bitcrush")?.at(1).bits).toBe(6);
+  });
+
+  it("takes a value that is not a number as the middle", () => {
+    // The knob has to land somewhere, and silence is the one answer that is
+    // never right.
+    expect(getAudioFxProfile("reverb")?.at(Number.NaN).size).toBe(
+      getAudioFxProfile("reverb")?.at(0.5).size,
+    );
+    expect(audioFxProfileStrength("reverb", {})).toBe(0.5);
+    expect(audioFxProfileStrength("not-an-effect", {})).toBe(0.5);
+  });
+});

packages/core/src/audioFxProfiles.ts

@@ -0,0 +1,209 @@
+/**
+ * One knob over several, for the five effects that cannot honestly have one
+ * real parameter nominated as the control that matters.
+ *
+ * A compressor has seven controls and an author wants one. The rest of the rack
+ * opens on a single real parameter — a filter's frequency, a delay's mix — but
+ * these five have no such parameter: threshold means nothing without ratio,
+ * ratio means nothing without make-up gain, and picking any one of them as the
+ * face of the module would be a knob that lies about what it sets.
+ *
+ * So they get a derived one, exactly as the carve already does: `carveProfile`
+ * turns a single 0..1 into six numbers, and these do the same for the rest of
+ * the rack. `EFFECT_COPY[id].primary` is `"strength"` for precisely these five,
+ * which is what says a module wants this treatment.
+ *
+ * Continuous rather than the three-point tables the design proposed. A table
+ * makes gentle/middle/strong three settings an author picks between, and the
+ * thing being modelled is not three settings — it is one axis, and the carve's
+ * knob has proved that reads. The proposal's figures survive as the anchors:
+ * each curve passes through them at 0, 0.5 and 1.
+ *
+ * Every number below is expressed in the parameter's own registry unit and then
+ * clamped by `normalizeAudioFxParams` on the way into the chain, so a profile
+ * cannot ship a value the effect would refuse.
+ */
+
+import { normalizeAudioFxParams, type HfAudioFxParamValues } from "./audioFx.js";
+
+/** 0..1, whatever arrives. NaN reads as the middle rather than as silence. */
+function clamp01(strength: number): number {
+  return Number.isFinite(strength) ? Math.min(1, Math.max(0, strength)) : 0.5;
+}
+
+/** Two decimal places, so a derived value reads as a setting and not as float noise. */
+function to2(value: number): number {
+  return Number(value.toFixed(2));
+}
+
+export interface HfAudioFxProfile {
+  /** What the one knob is called. Never the DSP name of anything it drives. */
+  label: string;
+  /** What the two ends sound like — the question an author actually has. */
+  ends: { low: string; high: string };
+  /** The parameters it sets. Everything else keeps the effect's own default. */
+  derives: readonly string[];
+  /** The mechanism values at a given strength. */
+  at(strength: number): HfAudioFxParamValues;
+}
+
+export const HF_AUDIO_FX_PROFILES: Record<string, HfAudioFxProfile> = {
+  compressor: {
+    label: "Evenness",
+    ends: { low: "Barely touched", high: "Very even, quite squashed" },
+    derives: ["threshold", "ratio", "attack", "release", "makeup"],
+    at(strength) {
+      const s = clamp01(strength);
+      return {
+        // Lower threshold and higher ratio together: more of the signal is
+        // caught, and what is caught is held harder. Moving one without the
+        // other is the pair of knobs this exists to stop an author meeting.
+        threshold: to2(-12 - s * 18),
+        ratio: to2(2 + s * 4),
+        // Faster as it gets firmer, because a firm compressor that lets peaks
+        // through is doing the audible half of its job and not the useful half.
+        attack: to2(25 - s * 20),
+        release: to2(300 - s * 210),
+        // Compression makes things quieter; this is the level put back, and it
+        // has to rise with the amount taken off or the knob reads as a volume
+        // control that goes the wrong way.
+        //
+        // Solved from measurement rather than proposed, and it is not linear:
+        // gain reduction accelerates as the threshold drops and the ratio rises
+        // together, so a straight line is too loud in the middle and too quiet
+        // at the top. The design's 1/3/7 dB left the track +0.9 dB at rest,
+        // +1.3 dB at the middle and −2.5 dB at full. Numbers in
+        // `~/audio-fx-profiles-ab/README.md`.
+        makeup: to2(s * s * 9.5),
+      };
+    },
+  },
+
+  gate: {
+    label: "Tightness",
+    ends: { low: "Only true silence", high: "Cuts quiet words too" },
+    derives: ["threshold", "range", "release"],
+    at(strength) {
+      const s = clamp01(strength);
+      return {
+        threshold: to2(-55 + s * 23),
+        // How far the gaps are ducked, never to silence: a room that stops dead
+        // between sentences sounds broken rather than clean.
+        range: to2(-10 - s * 20),
+        // Release has to come along, which the design did not have — and it
+        // dominates: swept against a fixed threshold and range, the gaps move
+        // 0.2 dB at 140 ms and 13.4 dB at 10 ms. Left at the effect's 100 ms
+        // default the gate was very nearly inaudible whatever else it was told.
+        //
+        // Measured on narration, speech level is unmoved (−24.0 dB) at every
+        // release down to 10 ms, so the usual reason to stay slow — clipping
+        // word endings — does not bite on this material at these thresholds.
+        release: to2(120 - s * 105),
+      };
+    },
+  },
+
+  saturate: {
+    label: "Warmth",
+    ends: { low: "Just a sheen", high: "Openly distorted" },
+    derives: ["threshold", "output"],
+    at(strength) {
+      const s = clamp01(strength);
+      return {
+        // Drive: the lower the threshold, the more of the signal meets the
+        // curve.
+        threshold: to2(-3 - s * 15),
+        // Up, not down — which reverses the design's figure, on the measurement
+        // that motivated taking one. A soft clipper at -18 dB threshold IS a
+        // limiter at -18 dB: peak fell to 0.089 from 0.496 and RMS almost
+        // halved, so the proposed trim of −3 dB made "warmer" mean "much
+        // quieter" and an author would have heard the level, not the warmth.
+        // Accelerating for the same reason as the compressor's make-up.
+        output: to2(s * s * 2.8),
+      };
+    },
+  },
+
+  reverb: {
+    label: "Space",
+    ends: { low: "A small tight room", high: "A big open hall" },
+    derives: ["size", "wet", "dry"],
+    at(strength) {
+      const s = clamp01(strength);
+      return {
+        // Anchored at the design's three figures — 0.25 / 0.55 / 0.90 — which a
+        // single linear run cannot hit, because 0.55 is not their midpoint.
+        size: to2(s <= 0.5 ? 0.25 + s * 0.6 : 0.55 + (s - 0.5) * 0.7),
+        wet: to2(0.15 + s * 0.3),
+        // Dry comes down as wet goes up, but not by the same amount: the two
+        // legs sum, and matching them exactly makes a big room quieter than a
+        // small one instead of further away.
+        dry: to2(0.92 - s * 0.2),
+      };
+    },
+  },
+
+  bitcrush: {
+    label: "Crush",
+    ends: { low: "Slightly gritty", high: "Destroyed" },
+    derives: ["bits", "samples", "mix"],
+    at(strength) {
+      const s = clamp01(strength);
+      return {
+        // Fewer steps and longer holds. `bits` runs DOWN as the knob runs up,
+        // which is why it cannot be the module's face on its own.
+        bits: to2(14 - s * 8),
+        samples: Math.max(1, Math.round(1 + s * 3)),
+        mix: to2(0.25 + s * 0.75),
+      };
+    },
+  },
+};
+
+export function getAudioFxProfile(type: string): HfAudioFxProfile | undefined {
+  return HF_AUDIO_FX_PROFILES[type];
+}
+
+/**
+ * The parameters a profile sets at this strength, merged over what is there.
+ *
+ * Merged rather than replacing: a profile names only the parameters it derives,
+ * and the rest — a compressor's knee, a saturation's curve type — are the
+ * author's to set under Details and must survive the knob moving.
+ */
+export function applyAudioFxProfile(
+  type: string,
+  strength: number,
+  params: HfAudioFxParamValues,
+): HfAudioFxParamValues {
+  const profile = getAudioFxProfile(type);
+  if (!profile) return params;
+  return normalizeAudioFxParams(type, { ...params, ...profile.at(strength) });
+}
+
+/**
+ * The strength a set of parameters reads as, by inverting the profile's own
+ * curve on its most characteristic parameter.
+ *
+ * A chain stores mechanism values, not the knob — the same contract
+ * `normalizeCarveSettings` has, and for the same reason: the mechanism is what
+ * renders, so it is what must be authoritative. Reading the knob back means one
+ * parameter has to be nominated as the one that says most about intent.
+ *
+ * A hand-edited chain therefore lands the knob at the nearest strength that
+ * would have produced its most telling value, which is the honest answer — the
+ * alternative is a knob parked at a default while the effect is set to
+ * something else entirely.
+ */
+export function audioFxProfileStrength(type: string, params: HfAudioFxParamValues): number {
+  const profile = getAudioFxProfile(type);
+  if (!profile) return 0.5;
+  const key = profile.derives[0];
+  if (key === undefined) return 0.5;
+  const value = params[key];
+  if (typeof value !== "number") return 0.5;
+  const low = profile.at(0)[key];
+  const high = profile.at(1)[key];
+  if (typeof low !== "number" || typeof high !== "number" || low === high) return 0.5;
+  return to2(Math.min(1, Math.max(0, (value - low) / (high - low))));
+}