Skip to content

Ears for coding agents: let AI hear, review and make sound. CPU-only, open source.

32.8 KB
Raw Blame

forge benchmark: v2r3

made: out/forge_v2; refs: /refsets.json; 2026-10-16 11:15

Medians per set. made = the made sounds vs the bench refs; real x = the forge refs (real, never in the bench) vs the bench refs; real loo = each bench ref vs the other bench refs. typ z, phrase z, margin, AS names: higher is better; tone dB, env oct (octaves off in attack / t40 / length), curve dB: lower is better; t40 x = tail length / bench median (1 = same). inside = checks (of n) a made sound lands inside the real 80% limit, median over the made sounds (real = the same count for the real rows). var cos = mean pairwise CLAP cosine within the set (made variants / forge refs / bench refs): high = the variants are near clones.

category set n typ z tone dB env oct curve dB t40 x phrase z margin AS names artifacts var cos inside
click made 5 0.21 8.79 0.23 8.83 1.44 3.03 0.24 0 0 0.723 7 / 7 (real 6)
real x 8 0.16 11.33 0.22 9.92 1.33 2.77 0.53 0.12 0 0.651
real loo 7 -0.1 9.85 0.46 13.19 0.86 2.67 0.27 0 0 0.576
downlifter made 6 3.23 0.92 0.33 0 0.749
note: empty: the licensed library has no downlifter, real or designed (no reversing or pitching of other sounds here: trim + fade only). Benchmark made downlifters on phrasings, AudioSet and artifacts only.
glitch made 6 2.03 -0.13 0 0 0.728
note: empty: no glitch material, real or designed, in the licensed library. Benchmark made glitches on phrasings, AudioSet and artifacts only.
hover_tick made 5 0.13 20.99 1.51 6.27 1.1 2.41 0.66 0 0 0.807 6 / 7 (real 5)
real x 6 0.09 21.2 1.64 9.88 0.9 2.51 0.9 0 0 0.799
real loo 6 0.47 15.46 1.16 19.39 1.2 2 0.58 0 4 0.626
impact made 6 5.31 2.47 30.34 0.56 3.34 0.88 1 0 0.905 6 / 6 (real 3)
real x 2 6.19 2.5 31.98 0.46 3.04 1.09 0.5 0 0.659
note: thin: the library has three cinematic hits (all three were the v1 impact forge refs); one is held out as bench, so v1 impact bench numbers are optimistic.
logo_sting made 5 16.35 1.22 14.81 1.03 1.32 -0.69 1 0 0.722 6 / 6 (real 5.5)
real x 2 15.52 1.46 21.27 1.7 1.4 -0.82 0.5 0 0.185
real loo 2 23.22 0.81 8.99 1.12 1.36 -0.58 1 0 0.507
note: no real logo stings exist in the library (they are composed); the refs are the struck metal / glass rings a sting ends on: a tail-shape and tone reference only.
notify made 6 0.15 18.29 0.81 13.55 0.85 1.4 -0.59 1 0 0.864 7 / 7 (real 5.5)
real x 5 -0.24 21.3 1.3 28.92 0.34 1.25 -0.24 1 0 0.593
real loo 5 3.84 16.7 0.81 17.41 1.19 1.76 -0.91 1 0 0.476
note: struck bells / glass / kalimba (forge) and celesta / glass / triangle (bench), rings capped at 2 s with a 0.25 s fade; no designed UI dings in the library.
pop made 5 -0.02 11.42 0.55 11.01 0.82 2.32 -0.2 0.4 0 0.735 7 / 7 (real 6)
real x 7 -0.09 16.06 1.51 20.54 0.91 2.46 -0.5 0.29 0 0.501
real loo 6 1.13 13.66 0.41 12.5 1.02 1.96 -1.06 0.67 0 0.522
pro_click made 5 15.43 1.28 13.47 0.8 3.91 2.28 0 0 0.786 5 / 6 (real 4)
real x 1 29.35 1.06 10.68 0.53 4.25 2.96 0 0
real loo 2 18.53 1.73 8.99 1.08 2.19 0.51 0 2 0.45
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md).
pro_glitch made 6 2.53 0.24 0 0 0.853
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md). forge only, one example.
pro_hover_tick made 5 -1.32 6.25 0.82 10.43 1 1.61 0.45 0 0 0.862 6 / 7 (real 6)
real x 3 -2.23 6.51 1.11 20.12 1 3.04 0.97 0 0 0.82
real loo 3 8.32 0.67 5.84 0.86 1.79 0.71 0 1 0.684
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md).
pro_impact made 5 2.55 -0.07 0.6 0 0.745
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md). forge only (one SFX-only film); designed, sub-heavy, gated tails.
pro_logo_sting made 5 24.43 2.32 23.36 0.13 0.83 -1.15 0.2 0 0.943 3 / 6 (real 6)
real x 1 15.49 2.84 24.51 0.06 0.99 -0.25 0 0
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md). forge: falling chirp into a tonal hum; bench: a music end chord with a 4 s tail.
pro_notify made 5 1.91 0.15 1 0 0.822
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md). forge only.
pro_pop made 6 23.63 1.41 35.59 0.75 3.32 1.81 0 0 0.686 3 / 6 (real 3)
real x 2 18.43 2.78 42.94 0.06 3.54 2.33 0 0 0.488
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md).
pro_pop_in made 5 0.93 -3.13 0 0 0.75
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md). not a forge category yet: element-in micro hits (thump + tick), forge only.
pro_riser made 5 17.88 1.54 36.57 0.05 2.71 0.47 0.2 0 0.712 4 / 6 (real 6)
real x 1 21.35 0.79 18.43 0.68 1.03 -0.72 0 0
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md). forge: one designed stutter riser; bench: one noise riser at a film end.
pro_shimmer made 5 3.29 1 0.2 0 0.646
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md). forge only; one granular sparkle + one melodic chime run.
pro_sub_drop made 5 24.84 0.54 16.11 1.64 2.44 0.68 0 0 0.835
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md). bench only, one example under music + voice.
pro_tap made 5 1.65 1.81 0.2 0 0.872
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md). forge only, one example.
pro_typing_key made 6 26.09 0.43 13.62 0.79 1.6 -0.96 0 0 0.815 4 / 6 (real 5.5)
real x 6 26.2 0.52 11.54 0.71 1.88 -0.42 0 0 0.849
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md).
pro_typing_run made 6 1.79 0.59 0 0 0.813
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md). not a forge category yet: one whole designed typing run.
pro_whoosh_big made 6 3.71 1.56 0.83 0 0.882
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md). forge only (one SFX-only film); the bench films bury their whooshes under music + voice.
pro_zip made 5 -2.41 11.89 0.71 20.81 0.84 2.34 -1.39 0 0 0.761 5 / 7 (real 5)
real loo 4 1.71 10.4 1 34.47 1.15 1.4 0.54 0 0 0.529
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md). not a forge category yet: tonal pitch sweeps, bench only (two films over music).
riser made 5 11.3 2.08 21.7 0.09 2.51 -0.3 0.4 0 0.82 3 / 6 (real 4)
real x 1 17.61 1.01 20.04 0.5 3.52 1.06 1 0
real loo 2 13.22 0.95 36.07 3.35 2.7 1.15 0.5 0 0.767
note: thin: one designed trailer riser (forge); the bench is two real cymbal swells, no synth risers.
shimmer made 6 1.07 10.68 0.35 4.06 0.97 3.46 0.91 0.67 0 0.845 7 / 7 (real 6)
real x 5 1.69 9.66 0.73 8.58 0.73 2.87 0.03 0.6 2 0.59
real loo 4 1.01 15.52 1.33 12.89 1.05 3.4 0.56 1 0 0.644
note: kit shimmer_01 (bar chimes sweep) is bench: its author also recorded the celesta notify bench.
sub_drop made 5 8.84 1.76 37.38 0.31 3.31 0.7 0 0 0.727
note: one designed source (a tom pitched down 24 semitones by its author), held out as bench; the forge has no sub-drop refs.
swipe made 5 -14.03 6.06 0.23 9.81 0.99 1.44 0.22 0.6 0 0.885 4 / 7 (real 6)
real x 4 -10.07 16.83 0.43 14.12 0.79 2.7 0.59 1 0 0.845
real loo 4 1.25 3.64 0.22 10.07 1.02 3.42 2.08 0.25 0 0.857
note: cloth only: the library has no paper slides.
swish made 5 -3.11 7.37 0.25 10.49 1.04 2.71 1.4 0 0 0.901 4 / 7 (real 6)
real x 7 1.94 11.08 0.64 12.88 0.64 3.22 1.6 0.57 0 0.811
real loo 3 9.01 0.91 16.45 1.16 3.5 1.92 0.67 0 0.79
tap made 5 -1.61 8.23 1.4 14.51 1.82 1.13 0.34 0.6 0 0.807 5 / 7 (real 6)
real x 7 -1.17 9.35 1.59 13.67 1.64 1.85 0.92 0.14 0 0.713
real loo 6 1.66 16.78 1.19 11.59 1 1.62 0.35 0.17 0 0.716
toggle made 5 -3.36 10.36 0.62 14.64 1.7 1.45 -1.35 0 0 0.718 5 / 7 (real 6)
real x 4 -2.84 12.48 1.87 18.78 2.11 2.1 -0.02 0 0 0.477
real loo 4 1.54 6.25 0.28 4.92 1.02 1.94 -0.28 0 0 0.725
typing_key made 6 -0.32 7.44 1.04 9.08 1 3.74 1.45 0 0 0.713 4 / 7 (real 6)
real x 15 0.26 8.88 1.01 12.2 1.29 4.75 2.83 0.53 0 0.721
real loo 12 0.23 5.47 1.52 16.7 1 3.71 1.53 0.42 0 0.673
whoosh_big made 6 0.77 6.67 0.44 8.97 0.97 2.38 0.03 0.83 0 0.854 7 / 7 (real 5)
real x 3 -0.27 8.8 1.13 19.46 3 2.92 0.04 1 0 0.7
real loo 3 4.25 0.43 6.67 0.97 1.88 0.19 0.33 0 0.495
note: forge = three designed Mixkit whooshes; bench = real air / cape swoops.
whoosh_fast made 6 -5.43 5.79 0.51 12.27 0.9 3.49 2.56 0.17 0 0.886 4.5 / 7 (real 6.5)
real x 5 0.57 10.31 1.22 20.2 0.6 4.14 2.78 0.4 0 0.826
real loo 5 1.52 4.58 0.4 8.21 1.08 4.14 2.62 0.8 0 0.772

Texture (medians; model-free, not in any forge objective): floor dB = 10th percentile 5 ms frame re the loudest (-120 = digital silence), stereo corr (1 = mono), darken = log2 centroid at half the tail / at the peak (< 0: the decay darkens).

category set floor dB stereo corr darken
click made -45.0 0.99 0.26
real x -42.85 0.99 -0.12
real loo -44.8 1.0 -0.07
downlifter made -54.45 0.67 -1.81
glitch made -44.95 0.09 0.04
hover_tick made -44.0 0.88 -0.09
real x -43.55 1.0 -0.09
real loo -24.9 0.52 -0.01
impact made -69.45 0.71 -1.21
real x -70.55 0.54 -1.3
logo_sting made -61.5 0.68 0.18
real x -52.1 0.59 -0.85
real loo -44.1 0.81 -1.43
notify made -44.75 0.98 -2.0
real x -41.2 0.99 -1.06
real loo -24.4 0.61 -0.53
pop made -62.8 1.0 0.25
real x -53.2 1.0 0.21
real loo -47.2 1.0 0.12
pro_click made -83.1 1.0
real x -108.5 0.0
real loo -30.6 0.95 0.26
pro_glitch made -44.85 0.04 0.04
pro_hover_tick made -37.8 0.26
real x -47.0 -0.0
real loo -37.4 0.16
pro_impact made -35.8 0.38 -0.25
pro_logo_sting made -60.2 1.0 -0.43
real x -120.0 1.0 -0.38
pro_notify made -47.7 0.94 0.39
pro_pop made -69.5 1.0 -1.45
real x -120.0 1.0
pro_pop_in made -61.5 0.58 -0.24
pro_riser made -32.5 0.82
real x -41.8 0.96 0.22
pro_shimmer made -49.0 0.33 -0.53
pro_sub_drop made -38.3 1.0 0.13
pro_tap made -80.8 0.44 -0.76
pro_typing_key made -38.95 1.0 0.02
real x -53.55 1.0 -0.04
pro_typing_run made -57.95 1.0
pro_whoosh_big made -55.05 1.0 -1.31
pro_zip made -52.9 0.99 -1.67
real loo -21.95 0.93 0.24
riser made -35.8 0.6 -0.03
real x -49.9 0.0
real loo -30.65 1.0 0.12
shimmer made -57.5 0.07 -0.64
real x -46.8 0.17 -0.29
real loo -36.1 0.68 -0.33
sub_drop made -67.0 1.0 -0.59
swipe made -65.2 0.9 -1.96
real x -56.85 0.42 0.07
real loo -38.05 0.68 0.61
swish made -61.5 0.94 -0.73
real x -55.6 0.87 -0.16
real loo -44.6 1.0 -0.45
tap made -45.0 0.93 0.0
real x -43.6 1.0 1.31
real loo -38.4 1.0 0.94
toggle made -59.4 0.99 0.45
real x -48.4 1.0 -0.11
real loo -49.1 0.48 -0.18
typing_key made -51.8 0.96 -0.43
real x -43.7 0.76 0.04
real loo -39.65 1.0 0.09
whoosh_big made -59.6 0.93 -1.2
real x -43.2 0.66 -0.11
real loo -42.9 1.0 0.28
whoosh_fast made -63.6 0.96 -0.64
real x -65.4 0.81 -0.16
real loo -32.2 1.0 -0.2

Diversity among the made variants (CLAP cosine mean / max, tonal distance min dB, near duplicates):

  • click: cos 0.723 / 0.837, tone min 7.41 dB, 0 near duplicate pair(s) (bench refs: cos 0.576 / 0.792, tone min 6.13 dB)
  • downlifter: cos 0.749 / 0.955, tone min 3.12 dB, 0 near duplicate pair(s)
  • glitch: cos 0.728 / 0.871, tone min 1.8 dB, 0 near duplicate pair(s)
  • hover_tick: cos 0.807 / 0.923, tone min 3.74 dB, 0 near duplicate pair(s) (bench refs: cos 0.626 / 0.847, tone min 9.21 dB)
  • impact: cos 0.905 / 0.967, tone min 2.22 dB, 0 near duplicate pair(s)
  • logo_sting: cos 0.722 / 0.876, tone min 10.32 dB, 0 near duplicate pair(s) (bench refs: cos 0.507 / 0.507, tone min 23.22 dB)
  • notify: cos 0.864 / 0.957, tone min 5.63 dB, 0 near duplicate pair(s) (bench refs: cos 0.476 / 0.905, tone min 11.33 dB)
  • pop: cos 0.735 / 0.877, tone min 5.92 dB, 0 near duplicate pair(s) (bench refs: cos 0.522 / 0.832, tone min 7.62 dB)
  • pro_click: cos 0.786 / 0.902, tone min 6.93 dB, 0 near duplicate pair(s) (bench refs: cos 0.45 / 0.45, tone min 18.53 dB)
  • pro_glitch: cos 0.853 / 0.912, tone min 1.35 dB, 0 near duplicate pair(s)
  • pro_hover_tick: cos 0.862 / 0.942, tone min 1.79 dB, 0 near duplicate pair(s) (bench refs: cos 0.684 / 0.776, tone min 6.36 dB)
  • pro_impact: cos 0.745 / 0.871, tone min 3.72 dB, 0 near duplicate pair(s)
  • pro_logo_sting: cos 0.943 / 0.976, tone min 1.26 dB, 0 near duplicate pair(s)
  • pro_notify: cos 0.822 / 0.902, tone min 6.02 dB, 0 near duplicate pair(s)
  • pro_pop: cos 0.686 / 0.876, tone min 7.51 dB, 0 near duplicate pair(s)
  • pro_pop_in: cos 0.75 / 0.823, tone min 5.62 dB, 0 near duplicate pair(s)
  • pro_riser: cos 0.712 / 0.808, tone min 2.42 dB, 0 near duplicate pair(s)
  • pro_shimmer: cos 0.646 / 0.808, tone min 7.57 dB, 0 near duplicate pair(s)
  • pro_sub_drop: cos 0.835 / 0.885, tone min 4.46 dB, 0 near duplicate pair(s)
  • pro_tap: cos 0.872 / 0.95, tone min 5.16 dB, 0 near duplicate pair(s)
  • pro_typing_key: cos 0.815 / 0.952, tone min 1.1 dB, 0 near duplicate pair(s)
  • pro_typing_run: cos 0.813 / 0.951, tone min 1.36 dB, 0 near duplicate pair(s)
  • pro_whoosh_big: cos 0.882 / 0.966, tone min 3.24 dB, 0 near duplicate pair(s)
  • pro_zip: cos 0.761 / 0.922, tone min 3.83 dB, 0 near duplicate pair(s) (bench refs: cos 0.529 / 0.686, tone min 8.63 dB)
  • riser: cos 0.82 / 0.88, tone min 4.7 dB, 0 near duplicate pair(s) (bench refs: cos 0.767 / 0.767, tone min 13.22 dB)
  • shimmer: cos 0.845 / 0.938, tone min 6.82 dB, 0 near duplicate pair(s) (bench refs: cos 0.644 / 0.765, tone min 10.66 dB)
  • sub_drop: cos 0.727 / 0.889, tone min 4.89 dB, 0 near duplicate pair(s)
  • swipe: cos 0.885 / 0.949, tone min 2.29 dB, 0 near duplicate pair(s) (bench refs: cos 0.857 / 0.918, tone min 2.76 dB)
  • swish: cos 0.901 / 0.957, tone min 4.36 dB, 0 near duplicate pair(s) (bench refs: cos 0.79 / 0.908, tone min 9.83 dB)
  • tap: cos 0.807 / 0.923, tone min 4.53 dB, 0 near duplicate pair(s) (bench refs: cos 0.716 / 0.98, tone min 2.68 dB)
  • toggle: cos 0.718 / 0.849, tone min 3.98 dB, 0 near duplicate pair(s) (bench refs: cos 0.725 / 0.846, tone min 5.33 dB)
  • typing_key: cos 0.713 / 0.885, tone min 6.37 dB, 0 near duplicate pair(s) (bench refs: cos 0.673 / 0.919, tone min 4.64 dB)
  • whoosh_big: cos 0.854 / 0.914, tone min 2.86 dB, 0 near duplicate pair(s) (bench refs: cos 0.495 / 0.768, tone min 3.17 dB)
  • whoosh_fast: cos 0.886 / 0.97, tone min 3.28 dB, 0 near duplicate pair(s) (bench refs: cos 0.772 / 0.951, tone min 2.28 dB)

Judge notes (round 3, integrator, 11:00-11:30)

Made = out/forge_v2/<category>/final/ as delivered after judging (183 sounds, 34 categories). Run: one full benchmark of the fixers' deliveries at 11:03 (kept as out/forge_v2/_judge/results_v2r3_asdelivered.json, not committed), then the same run again at 11:15 after the judge put the round-2 finals back for two categories (this file; the other 32 categories are byte-identical between the two runs and give the same numbers). Per-area fixers had each run the bench once at the end of their own work (the ui area 4 times, see its report); the judge read bench numbers only to decide r2 vs r3, never to tune. Model-free columns (band shares, onset brightness, t40 ms, stereo corr) are the judge's own measurements (out/forge_v2/_judge/judge_r3.py, sheets out/forge_v2/_judge/r3/<category>.png: r2 / r3 / forge-ref spectrograms, envelopes, 1/3-octave profiles with the bench median as a dashed line; not committed). Nobody has listened yet.

Round 2 -> round 3 (made medians on the held-out bench; v2r2 vs v2r3)

real = the same count for real recordings; real x = forge refs vs bench; loo = bench vs bench; var cos lower = more varied, (bench) = the bench refs' own; AS names = share of sounds whose AudioSet top 3 names the category. Named fault = the round-2 judge's fault (or the fixer's target) measured model-free, r2 -> r3 (reference median); band shares are dB of total energy, onset >6 kHz = share of the first 15 ms.

category n inside (real) tone dB (real x) t40 x (real x / loo) var cos (bench) phrase z AS names named fault, model-free (r2 -> r3, ref) artifacts
click 5 -> 5 5 -> 7 /7 (6) 9.86 -> 8.79 (11.33) 1.22 -> 1.44 (1.33 / 0.86) 0.681 -> 0.723 (0.576) 2.27 -> 3.03 0 -> 0 1-4 kHz share -14 -> -5.3 (bench -4.9) 0 -> 0
downlifter 6 -> 6 - - - 0.749 -> 0.749 (-) 3.23 -> 3.23 0.33 -> 0.33 - 0 -> 0
glitch 4 -> 6 - - - 0.799 -> 0.728 (-) 2.14 -> 2.03 0.50 -> 0 - 0 -> 0
hover_tick 5 -> 5 6 -> 6 /7 (5) 20.99 -> 20.99 (21.20) 1.10 -> 1.10 (0.90 / 1.20) 0.807 -> 0.807 (0.626) 2.41 -> 2.41 0 -> 0 - 0 -> 0
impact 5 -> 6 5 -> 6 /6 (3) 6.29 -> 5.31 (6.19) 0.62 -> 0.56 (0.46 / -) 0.881 -> 0.905 (-) 3.61 -> 3.34 1 -> 1 onset >6 kHz (15 ms) -28.9 -> -10.9 (bench -21.6) 1 -> 0
logo_sting 4 -> 5 3.5 -> 6 /6 (5.5) 16.29 -> 16.35 (15.52) 2.48 -> 1.03 (1.70 / 1.12) 0.733 -> 0.722 (0.507) 1.55 -> 1.32 1 -> 1 t40 ms 2135 -> 885 (bench 862.5) 1 -> 0
notify 5 -> 6 7 -> 7 /7 (5.5) 18.65 -> 18.29 (21.30) 0.63 -> 0.85 (0.34 / 1.19) 0.810 -> 0.864 (0.476) 2.21 -> 1.40 1 -> 1 150 Hz-1 kHz -41.3 -> -1 (bench 0) 0 -> 0
pop 5 -> 5 7 -> 7 /7 (6) 11.42 -> 11.42 (16.06) 0.82 -> 0.82 (0.91 / 1.02) 0.735 -> 0.735 (0.522) 2.32 -> 2.32 0.40 -> 0.40 1-4 kHz -17.2 -> -17.2 (bench -2) 0 -> 0
pro_click 5 -> 5 4 -> 5 /6 (4) 25.79 -> 15.43 (29.35) 0.67 -> 0.80 (0.53 / 1.08) 0.870 -> 0.786 (0.450) 4.06 -> 3.91 0 -> 0 150 Hz-1 kHz -41.2 -> -6.8 (bench -6.2) 0 -> 0
pro_glitch 5 -> 6 - - - 0.875 -> 0.853 (-) 3.02 -> 2.53 0.60 -> 0 <150 Hz -39.9 -> -21.9 (forge -21.6) 0 -> 0
pro_hover_tick 5 -> 5 4 -> 6 /7 (6) 7.89 -> 6.25 (6.51) 1.17 -> 1 (1 / 0.86) 0.878 -> 0.862 (0.684) 0.85 -> 1.61 0 -> 0 stereo corr 1 -> 0.3 (bench 0.2) 0 -> 0
pro_impact 5 -> 5 - - - 0.733 -> 0.745 (-) 2.71 -> 2.55 0.80 -> 0.60 4-12 kHz -30.4 -> -17.7 (forge -15.4) 0 -> 0
pro_logo_sting 4 -> 5 3 -> 3 /6 (6) 24.42 -> 24.43 (15.49) 0.15 -> 0.13 (0.06 / -) 0.947 -> 0.943 (-) 0.73 -> 0.83 0 -> 0.20 1-4 kHz -13.3 -> -11 (bench -9.2) 0 -> 0
pro_notify 5 -> 5 - - - 0.882 -> 0.822 (-) 1.71 -> 1.91 1 -> 1 >12 kHz -29.6 -> -19.6 (forge -17.3) 0 -> 0
pro_pop 5 -> 6 4 -> 3 /6 (3) 23.68 -> 23.63 (18.43) 0.81 -> 0.75 (0.06 / -) 0.663 -> 0.686 (-) 3.29 -> 3.32 0 -> 0 1-4 kHz -3.1 -> -1 (bench -2.4) 0 -> 0
pro_pop_in 5 -> 5 - - - 0.895 -> 0.750 (-) 0.62 -> 0.93 0 -> 0 1-4 kHz -29.6 -> -24.7 (forge -27.5) 0 -> 0
pro_riser 5 -> 5 3 -> 4 /6 (6) 20.80 -> 17.88 (21.35) 0.05 -> 0.05 (0.68 / -) 0.676 -> 0.712 (-) 2 -> 2.71 0 -> 0.20 <150 Hz -9.4 -> -24.1 (bench -42.5) 0 -> 0
pro_shimmer 5 -> 5 - - - 0.646 -> 0.646 (-) 3.29 -> 3.29 0.20 -> 0.20 - 0 -> 0
pro_sub_drop 5 -> 5 - 24.84 -> 24.84 (-) 1.64 -> 1.64 (- / -) 0.835 -> 0.835 (-) 2.44 -> 2.44 0 -> 0 - 0 -> 0
pro_tap 5 -> 5 - - - 0.854 -> 0.872 (-) 1.73 -> 1.65 0 -> 0.20 >12 kHz -57.8 -> -47.9 (forge -36.6) 0 -> 0
pro_typing_key 5 -> 6 4 -> 4 /6 (5.5) 26.38 -> 26.09 (26.20) 0.64 -> 0.79 (0.71 / -) 0.926 -> 0.815 (-) 1.84 -> 1.60 0 -> 0 - 0 -> 0
pro_typing_run 5 -> 6 - - - 0.930 -> 0.813 (-) 2 -> 1.79 0 -> 0 - 1 -> 0
pro_whoosh_big 5 -> 6 - - - 0.931 -> 0.882 (-) 3.93 -> 3.71 1 -> 0.83 1-4 kHz -46.5 -> -34.8 (forge -38); 4-12 kHz -31.7 -> -41.7 (forge -38.5) 0 -> 0
pro_zip 5 -> 5 5 -> 5 /7 (5) 11.89 -> 11.89 (-) 0.84 -> 0.84 (- / 1.15) 0.761 -> 0.761 (0.529) 2.34 -> 2.34 0 -> 0 4-12 kHz -13.1 -> -13.1 (bench -16.1) 0 -> 0
riser 5 -> 5 3 -> 3 /6 (4) 12.50 -> 11.30 (17.61) 0.06 -> 0.09 (0.50 / 3.35) 0.780 -> 0.820 (0.767) 2.57 -> 2.51 0.80 -> 0.40 <150 Hz -16.2 -> -19.6 (bench -28.1) 0 -> 0
shimmer 5 -> 6 6 -> 7 /7 (6) 10.49 -> 10.68 (9.66) 0.48 -> 0.97 (0.73 / 1.05) 0.807 -> 0.845 (0.644) 3.42 -> 3.46 0.40 -> 0.67 t40 ms 1155 -> 2340 (bench 2420) 0 -> 0
sub_drop 5 -> 5 - 8.84 -> 8.84 (-) 0.31 -> 0.31 (- / -) 0.727 -> 0.727 (-) 3.31 -> 3.31 0 -> 0 - 0 -> 0
swipe 5 -> 5 4 -> 4 /7 (6) 13.47 -> 6.06 (16.83) 0.86 -> 0.99 (0.79 / 1.02) 0.855 -> 0.885 (0.857) 1.82 -> 1.44 0.40 -> 0.60 4-12 kHz share -35.9 -> -20.7 (bench -19.2); onset >6 kHz -27.4 -> -22.4 (bench -11.6) 0 -> 0
swish 5 -> 5 4 -> 4 /7 (6) 7.48 -> 7.37 (11.08) 0.68 -> 1.04 (0.64 / 1.16) 0.860 -> 0.901 (0.790) 2.68 -> 2.71 0 -> 0 t40 ms 150 -> 230 (bench 220); <150 Hz -15.8 -> -23.4 (bench -17.4); 4-12 kHz -18.2 -> -29.1 (bench -33.5) 0 -> 0
tap 5 -> 5 3 -> 5 /7 (6) 6.10 -> 8.23 (9.35) 1.27 -> 1.82 (1.64 / 1) 0.856 -> 0.807 (0.716) 0.52 -> 1.13 0 -> 0.60 4-12 kHz -27.5 -> -18.4 (bench -26.4) 0 -> 0
toggle 5 -> 5 5 -> 5 /7 (6) 8.82 -> 10.36 (12.48) 1.26 -> 1.70 (2.11 / 1.02) 0.720 -> 0.718 (0.725) 1.39 -> 1.45 0 -> 0 150 Hz-1 kHz -5.4 -> -9.8 (bench -19.2) 0 -> 0
typing_key 5 -> 6 6 -> 4 /7 (6) 8.15 -> 7.44 (8.88) 1.29 -> 1 (1.29 / 1.00) 0.705 -> 0.713 (0.673) 4.46 -> 3.74 0.20 -> 0 stereo corr 0.3 -> 1 (bench 1); <150 Hz -5.5 -> -10.2 (bench -10.9) 0 -> 0
whoosh_big 5 -> 6 7 -> 7 /7 (5) 7.58 -> 6.67 (8.80) 1.04 -> 0.97 (3 / 0.97) 0.909 -> 0.854 (0.495) 2.58 -> 2.38 1 -> 0.83 - 0 -> 0
whoosh_fast 5 -> 6 4 -> 4.5 /7 (6.5) 9.43 -> 5.79 (10.31) 0.49 -> 0.90 (0.60 / 1.08) 0.856 -> 0.886 (0.772) 3.28 -> 3.49 0.60 -> 0.17 t40 ms 130 -> 237.5 (bench 265); <150 Hz -9.7 -> -14.4 (bench -17.7); 4-12 kHz -17.4 -> -26.4 (bench -32.3) 0 -> 0

Round-3 attempts the judge did NOT deliver (kept in <category>/final_r3/, round-2 finals back in final/):

category n inside tone dB t40 x env oct curve dB phrase z var cos why rejected
pop 5 7 /7 (r2 7) 12.38 (r2 11.42) 0.55 (r2 0.82; real x 0.91) 0.66 (0.55) 11.43 (11.01) 1.83 (2.32) 0.665 (0.735) worse on every single-sound check that moved; the snap adds 4-12 kHz (-25.7 dB vs bench -35.5) and shortens the tail
pro_zip 6 4.5 /7 (r2 5) 12.72 (r2 11.89) 0.63 (r2 0.84; loo 1.15) 0.97 (0.71) 22.95 (20.81) 1.52 (2.34) 0.768 (0.761) worse on all four model-free checks; its one gain (margin -1.39 -> -0.74) came after tuning on an AudioSet reading

Verdicts per category (good for a motion-design launch video?)

Fixed as asked (the named fault measurably moved to the reference, no new fault found):

  • swipe: muffle fixed. 4-12 kHz -35.9 -> -20.7 dB (bench cloth -19.2), tone 13.47 -> 6.06 dB (real x 16.83), t40 x 0.99. Onset still 11 dB duller than the bench cloth (>6 kHz -22.4 vs -11.6); L/R corr 0.90 vs 0.70; AudioSet 'Door' on 1 of 5.
  • whoosh_fast / swish: tails x1.8 (t40 130 -> 238 / 150 -> 230 ms; bench 265 / 220), boom down 5-8 dB, hiss down 9-11 dB; whoosh_fast tone 9.43 -> 5.79 dB (real x 10.31), curve 18.4 -> 12.3. Every sound is inside the real limits on the three model-free checks; the misses are CLAP / AudioSet (AudioSet still hears Door / Knock / Slam first on 6 of 6 and 5 of 5), and variety got worse (0.856 -> 0.886, 0.860 -> 0.901). Still under v1's 7/7 inside (v1 sat above the real rows on the saturated model checks, tuned on kit refs that are bench refs now).
  • logo_sting: redefined as a composed reveal (lead-in -> strike -> one note or dyad -> short tail). t40 2135 -> 885 ms (bench 862), curve 25.2 -> 14.8 dB, inside 3.5 -> 6 of 6 (real 5.5), artifact gone, AudioSet 'Ding' first on 5 of 5. New weakness seen on the sheet: no body at all (<150 Hz -52.7 dB of total vs bench -26.8; nothing under ~800 Hz).
  • typing_key: width fixed (stereo corr 0.34 -> 0.96, bench 1.00), thock down (<150 Hz -5.5 -> -10.2, bench -10.9), tone 7.44, t40 x 1.00, curve 9.1; the model checks fell (inside 6 -> 4 of 7, phrase z 4.46 -> 3.74): CLAP hears less low end, not a fault on the trust order.
  • click: 1-4 kHz -14.0 -> -5.3 dB (bench -4.9), late partial line 18.7 -> 15.0 dB (refs 15.1, fixer's measure), inside 5 -> 7 of 7, phrase z 2.27 -> 3.03. Variety slipped 0.681 -> 0.723 (bench 0.576).
  • shimmer: rings out freely: t40 1155 -> 2340 ms (bench 2420), curve 14.1 -> 4.1 dB, inside 7 of 7, AudioSet 0.4 -> 0.67.
  • pro_hover_tick: spike + grains: phrase z 0.85 -> 1.61 (loo 1.79), margin -0.54 -> 0.45, tone 6.25 (real x 6.51), inside 6 of 7.
  • pro_click: bodies added: 150 Hz-1 kHz -41.2 -> -6.8 dB (bench -6.2), tone 25.8 -> 15.4 dB, inside 5 of 6, var 0.786.
  • pro_glitch: burst drama: median 2 ms level re peak -22 -> -36 dB (judge's own measure; forge film -33), the stutter speeds up in 6 of 6, <150 Hz -21.9 (film -21.6). AudioSet no longer says static (0.6 -> 0), phrase z 3.02 -> 2.53.
  • pro_typing_key / pro_typing_run / pro_pop_in: clones fixed: var cos 0.926 -> 0.815, 0.930 -> 0.813, 0.895 -> 0.750 (film sets 0.70-0.85); the run's tail-cut flag is gone. pro_typing_key 06 'crackle' is outside the study's envelope box.
  • whoosh_big / pro_whoosh_big: variety 0.909 -> 0.854 and 0.931 -> 0.882; whoosh_big still inside 7 of 7 (real 5), tone 7.58 -> 6.67. Costs: whoosh_big curve 5.8 -> 9.0 dB, AudioSet names 5 of 5 -> 5 of 6. pro_whoosh_big's air layer is fused (1-4 kHz -46.5 -> -34.8, film -38.0).
  • pro_riser: tone 20.8 -> 17.9 dB, phrase z 2.00 -> 2.71, margin -0.78 -> 0.47, inside 3 -> 4 of 6; the gaps dip 10-20 dB where the film's drop 40; AudioSet hears 'Fart' / 'Machine gun' on two members.

Better with a cost (delivered, the cost noted):

  • impact: onset fixed and then overshot: >6 kHz share of the first 15 ms -28.9 -> -10.9 dB (bench -21.6; the >12 kHz share of that window is -10..-14 dB vs bench -28, forge -33..-45), so the contact now carries a bright tick; after 100 ms the air is back at the r2 / bench level. Tone 6.29 -> 5.31 dB (better than real x 6.19), inside 5 -> 6 of 6, late-start flag gone; variety worse 0.881 -> 0.905. Next step: the sheen 8-10 dB down.
  • toggle: envelope fixed (env oct 2.58 -> 0.62, curve 25.4 -> 14.6: the first contact is now the loudest), but tone 8.82 -> 10.36 dB, t40 x 1.26 -> 1.70 and margin -0.32 -> -1.35 got worse; toggle_01 / 03 show narrow ringing lines at 3-10 kHz.
  • tap: reads more like a tap (phrase z 0.52 -> 1.13, AudioSet 0 -> 0.6, inside 3 -> 5 of 7) but is 8 dB brighter at 4-12 kHz than the bench (-18.4 vs -26.4), tone 6.10 -> 8.23, tails long (t40 x 1.82).
  • notify: lower, longer bells (150 Hz-1 kHz -41 -> -1 dB, bench 0; t40 x 0.63 -> 0.85; curve 21.8 -> 13.6), but phrase z 2.21 -> 1.40 and AudioSet 'Clang' first on 4 of 6 (r2 'Ding'); variety 0.81 -> 0.86.
  • pro_impact: brighter grains toward the film (4-12 kHz -30.4 -> -17.7, film -15.4), straight dB fall before the gate; the grains are evenly spaced stripes (as in one of the film refs) and margin fell 0.60 -> -0.07.
  • pro_notify: cleaner head, breath and >12 kHz toward the films (-29.6 -> -19.6, films -17.3), var 0.882 -> 0.822; still far under the films on today's critic (-5.97 vs +0.1..3.0, fixer's rescore).
  • riser: tone 12.5 -> 11.3, curve 24.8 -> 21.7, longest steady line 0.22 -> 0.11 s (fixer); AudioSet names 0.8 -> 0.4.
  • glitch: 6 members, variety 0.799 -> 0.728; AudioSet names 0.5 -> 0, margin 0.08 -> -0.13 (no refs: designed mode only).
  • pro_logo_sting: more weight (1-4 kHz -13.3 -> -11.0) but still near clones (0.943) and band-limited: nothing above ~5 kHz (4-12 kHz -41.8 dB of total vs the film -21.0).
  • pro_pop / pro_tap: roughly flat (pro_pop inside 4 -> 3 of 6 = real 3; pro_tap phrase z 1.73 -> 1.65, variety 0.854 -> 0.872).

Round 2 kept: pop, pro_zip (above). Unchanged in round 3: downlifter, hover_tick, sub_drop, pro_sub_drop, pro_shimmer.

Showcase-ready (measured, not yet listened): whoosh_big, pro_whoosh_big, shimmer, click, typing_key, pro_typing_key, pro_typing_run, pro_pop_in, pro_hover_tick, pro_click, logo_sting, swipe, impact (mind the contact tick), whoosh_fast and swish (tail and balance fixed; AudioSet still says door), pop (r2), pro_zip (r2), downlifter, sub_drop, pro_glitch, pro_riser. Weakest now: pro_logo_sting (clones, no top end), pro_tap, pro_pop, toggle (ring, tone), tap (bright, long), notify ('Clang'), riser (AudioSet), glitch (no refs), pro_notify (far under the films on the critic).

Critic passes 55 / 183 (r2 60 / 168), not comparable: the core critic changed during round 3 and the areas used different critic setups (see each LOG). Tests: full fast suite 400 passed, 15 skipped (316 s).

Pages for the morning (private: they play licensed forge refs): out/showcase/gallery.html (round-3 finals with these numbers) and out/showcase/compare.html (r2 / r3 / r3x rows per category).