Skip to content

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

27.8 KB
Raw Blame

forge benchmark: v2r2

made: out/forge_v2; refs: /refsets.json; 2026-10-16 08:44

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 -0.25 9.86 0.24 8.61 1.22 2.27 0.56 0 0 0.681 5 / 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 4 2.14 0.08 0.5 0 0.799
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 5 6.29 2.11 32.62 0.62 3.61 1.21 1 1 0.881 5 / 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 4 16.29 2.01 25.16 2.48 1.55 -0.85 1 1 0.733 3.5 / 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 5 -0.02 18.65 0.57 21.83 0.63 2.21 -0.65 1 0 0.81 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 25.79 1.46 13.82 0.67 4.06 2.64 0 0 0.87 4 / 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 5 3.02 0.95 0.6 0 0.875
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 -1.81 7.89 0.88 10.7 1.17 0.85 -0.54 0 0 0.878 4 / 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 2.71 0.6 0.8 0 0.733
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 4 24.42 2.33 22.75 0.15 0.73 -1.3 0 0 0.947 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 1.71 0.08 1 0 0.882
note: pro designed SFX from the reference launch films (bed under some trims; see docs/motion-sfx-study.md). forge only.
pro_pop made 5 23.68 1.25 37.49 0.81 3.29 1.89 0 0 0.663 4 / 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 0.62 -3.53 0 0 0.895
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 20.8 1.61 36.34 0.05 2 -0.78 0 0 0.676 3 / 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 1.73 1.9 0 0 0.854
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 5 26.38 0.51 15.74 0.64 1.84 -0.85 0 0 0.926 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 5 2 0.77 0 1 0.93
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 5 3.93 1.6 1 0 0.931
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 12.5 2.12 24.81 0.06 2.57 -0.28 0.8 0 0.78 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 5 0.12 10.49 0.99 14.09 0.48 3.42 0.7 0.4 0 0.807 6 / 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 -14.89 13.47 0.34 6.68 0.86 1.82 0.34 0.4 0 0.855 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 -2.8 7.48 0.64 13.78 0.68 2.68 1.62 0 0 0.86 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 -2.69 6.1 1.05 13.46 1.27 0.52 0.13 0 0 0.856 3 / 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 -3.59 8.82 2.58 25.45 1.26 1.39 -0.32 0 0 0.72 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 5 -0.15 8.15 0.97 11.12 1.29 4.46 2.26 0.2 0 0.705 6 / 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 5 0.78 7.58 0.41 5.82 1.04 2.58 0.13 1 0 0.909 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 5 -4.85 9.43 0.95 18.4 0.49 3.28 2.46 0.6 0 0.856 4 / 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 -45.2 1.0 -0.12
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 -39.5 0.13 0.01
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 -71.8 0.74 -0.8
real x -70.55 0.54 -1.3
logo_sting made -44.75 0.69 -0.39
real x -52.1 0.59 -0.85
real loo -44.1 0.81 -1.43
notify made -28.6 0.9 -1.01
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 -75.2 1.0
real x -108.5 0.0
real loo -30.6 0.95 0.26
pro_glitch made -36.1 0.04 0.02
pro_hover_tick made -41.8 0.97 -0.14
real x -47.0 -0.0
real loo -37.4 0.16
pro_impact made -24.4 0.59 0.45
pro_logo_sting made -55.25 1.0 0.2
real x -120.0 1.0 -0.38
pro_notify made -45.7 0.94 -0.27
pro_pop made -96.9 1.0 -1.9
real x -120.0 1.0
pro_pop_in made -87.2 0.58 -0.89
pro_riser made -29.9 0.91
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 -101.6 0.46 -0.27
pro_typing_key made -42.7 1.0 0.02
real x -53.55 1.0 -0.04
pro_typing_run made -56.4 1.0
pro_whoosh_big made -59.1 1.0 -0.87
pro_zip made -52.9 0.99 -1.67
real loo -21.95 0.93 0.24
riser made -33.4 0.61
real x -49.9 0.0
real loo -30.65 1.0 0.12
shimmer made -53.9 0.33 -0.16
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 -60.5 0.93 -1.27
real x -56.85 0.42 0.07
real loo -38.05 0.68 0.61
swish made -63.0 0.95 -1.15
real x -55.6 0.87 -0.16
real loo -44.6 1.0 -0.45
tap made -65.9 0.98 0.07
real x -43.6 1.0 1.31
real loo -38.4 1.0 0.94
toggle made -49.5 0.99 -0.48
real x -48.4 1.0 -0.11
real loo -49.1 0.48 -0.18
typing_key made -49.0 0.34 -1.17
real x -43.7 0.76 0.04
real loo -39.65 1.0 0.09
whoosh_big made -61.6 0.95 -1.04
real x -43.2 0.66 -0.11
real loo -42.9 1.0 0.28
whoosh_fast made -64.4 0.98 -1.14
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 0.681 / 0.83, tone min 6.69 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 0.799 / 0.88, tone min 5.31 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 0.881 / 0.966, tone min 2.34 dB, 0 near duplicate pair(s)
  • logo_sting: cos 0.733 / 0.901, tone min 9.37 dB, 0 near duplicate pair(s) (bench refs: cos 0.507 / 0.507, tone min 23.22 dB)
  • notify: cos 0.81 / 0.92, tone min 7.59 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 0.87 / 0.904, tone min 3.66 dB, 0 near duplicate pair(s) (bench refs: cos 0.45 / 0.45, tone min 18.53 dB)
  • pro_glitch: cos 0.875 / 0.94, tone min 2.69 dB, 0 near duplicate pair(s)
  • pro_hover_tick: cos 0.878 / 0.949, tone min 3.83 dB, 0 near duplicate pair(s) (bench refs: cos 0.684 / 0.776, tone min 6.36 dB)
  • pro_impact: cos 0.733 / 0.827, tone min 4.73 dB, 0 near duplicate pair(s)
  • pro_logo_sting: cos 0.947 / 0.968, tone min 2.38 dB, 0 near duplicate pair(s)
  • pro_notify: cos 0.882 / 0.947, tone min 5.93 dB, 0 near duplicate pair(s)
  • pro_pop: cos 0.663 / 0.849, tone min 9.48 dB, 0 near duplicate pair(s)
  • pro_pop_in: cos 0.895 / 0.942, tone min 5.0 dB, 0 near duplicate pair(s)
  • pro_riser: cos 0.676 / 0.802, tone min 5.68 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 0.854 / 0.892, tone min 7.57 dB, 0 near duplicate pair(s)
  • pro_typing_key: cos 0.926 / 0.966, tone min 1.06 dB, 0 near duplicate pair(s)
  • pro_typing_run: cos 0.93 / 0.958, tone min 1.36 dB, 0 near duplicate pair(s)
  • pro_whoosh_big: cos 0.931 / 0.954, tone min 3.66 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 0.78 / 0.887, tone min 5.02 dB, 0 near duplicate pair(s) (bench refs: cos 0.767 / 0.767, tone min 13.22 dB)
  • shimmer: cos 0.807 / 0.907, tone min 8.08 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 0.855 / 0.924, tone min 2.77 dB, 0 near duplicate pair(s) (bench refs: cos 0.857 / 0.918, tone min 2.76 dB)
  • swish: cos 0.86 / 0.942, tone min 3.97 dB, 0 near duplicate pair(s) (bench refs: cos 0.79 / 0.908, tone min 9.83 dB)
  • tap: cos 0.856 / 0.892, tone min 5.15 dB, 0 near duplicate pair(s) (bench refs: cos 0.716 / 0.98, tone min 2.68 dB)
  • toggle: cos 0.72 / 0.877, tone min 4.1 dB, 0 near duplicate pair(s) (bench refs: cos 0.725 / 0.846, tone min 5.33 dB)
  • typing_key: cos 0.705 / 0.87, tone min 6.24 dB, 0 near duplicate pair(s) (bench refs: cos 0.673 / 0.919, tone min 4.64 dB)
  • whoosh_big: cos 0.909 / 0.96, tone min 3.02 dB, 0 near duplicate pair(s) (bench refs: cos 0.495 / 0.768, tone min 3.17 dB)
  • whoosh_fast: cos 0.856 / 0.931, tone min 3.86 dB, 0 near duplicate pair(s) (bench refs: cos 0.772 / 0.951, tone min 2.28 dB)

Judge notes (round 2, integrator)

Made = out/forge_v2/<category>/final/<category>_NN.wav (168 sounds, 34 categories; audition*.wav skipped). Everything below comes from this run, eval/forge/results_v1.json (v1 finals, same frozen benchmark), the critic reports in each final/report.json, and spectrogram / envelope / 1/3-octave sheets of finals vs bench, forge and pro refs (out/forge_v2/_judge/<category>.png, not committed). Band shares are dB of total energy (frozen_tone.shares), medians per set. Nobody has listened yet: final/audition.wav per category is ready for that.

v1 vs v2 round 2 (the five categories v1 covered; made medians on the held-out bench)

category var cos v1 -> v2 (real) t40 x v1 -> v2 (real x / loo) tone dB v1 -> v2 (real x / loo) env oct curve dB typ z AS names inside v1 -> v2 (real)
click 0.886 -> 0.681 (0.65 / 0.58) 0.56 -> 1.22 (1.33 / 0.86) 7.60 -> 9.86 (11.33 / 9.85) 0.56 -> 0.24 8.85 -> 8.61 1.89 -> -0.25 0 -> 0 7 -> 5 of 7 (6)
typing_key 0.965 -> 0.705 (0.72 / 0.67) 1.67 -> 1.29 (1.29 / 1.00) 8.20 -> 8.15 (8.88 / 5.47) 1.07 -> 0.97 14.32 -> 11.12 1.20 -> -0.15 0.33 -> 0.20 7 -> 6 of 7 (6)
pop 0.917 -> 0.735 (0.50 / 0.52) 1.00 -> 0.82 (0.91 / 1.02) 8.59 -> 11.42 (16.06 / 13.66) 0.68 -> 0.55 6.29 -> 11.01 1.33 -> -0.02 1.0 -> 0.4 7 -> 7 of 7 (6)
impact 0.931 -> 0.881 (0.66) 0.50 -> 0.62 (0.46) 6.70 -> 6.29 (6.19) 1.54 -> 2.11 24.86 -> 32.62 - 1.0 -> 1.0 5 -> 5 of 6 (3)
whoosh_fast 0.957 -> 0.856 (0.83 / 0.77) 0.51 -> 0.49 (0.60 / 1.08) 8.58 -> 9.43 (10.31 / 4.58) 0.89 -> 0.95 15.93 -> 18.40 2.05 -> -4.85 0.33 -> 0.6 7 -> 4 of 7 (6.5)

Read: the round-1 clone finding is fixed in all five (var cos now at or near the real sets), the click and key tails are fixed (t40 x at the real level), and typ z / phrase z dropped from "above real" (prototypes) to real-like. The fast whoosh is the exception: its tail is still half the bench median and it now misses 3 of 7 real limits that v1 met. "inside" falls for click because v1 sat above the real rows on the saturated model checks.

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

Critic passes count the final members passing their own category critic (different refs per category, so totals are not comparable across categories). 60 of 168 finals pass.

category verdict key numbers (bench, made median) concrete fault
whoosh_big best inside 7/7 (real 5), tone 7.58 (real x 8.80), t40 x 1.04, curve 5.8 dB, AudioSet "Whoosh" 5/5, critic 5/5 variants near clones (var cos 0.909 vs bench 0.495)
typing_key good inside 6/7 (real 6), var 0.705, t40 x 1.29 = real x, curve 11.1 (real x 12.2), critic 3/5 too wide (stereo corr 0.34, bench 1.00); <150 Hz share -5.5 vs -10.9 dB (thock boom); AudioSet names it 0.2
pro_pop_in good 1/3-oct balance and envelopes overlay the 10 film refs, double hits 60-70 ms, 0 artifacts near clones (0.895 vs refs 0.696); critic 0/5 (CLAP hears every low thump as a kick drum)
pro_whoosh_big good critic 5/5, AudioSet "Whoosh" 4/5, <150 Hz share -0.3 vs refs -0.2 detached 5-15 kHz air island over a 1-4 kHz hole (1-4k share -46.5 vs -38.0); near clones 0.931
pro_typing_key / pro_typing_run good balance overlays the 6 film keys; run 01 matches the film's spacing (87 ms median); critic 1/5, 2/5 near clones 0.926 / 0.930 (the film keys themselves 0.849)
downlifter good (no refs) critic 6/6, var 0.749, 0 artifacts, phrase z 3.23 judged on models + gesture only; 06 sub_sink reads "Bang" (0.51)
pro_zip good inside 5/7 (real 5), tone 11.9 (bench loo 10.4), var 0.761, critic 5/5 AudioSet "Boing" / "Plop" on 2 of 5: toy-like pure tone
sub_drop / pro_sub_drop good critic 4/5, 3/5; 0 artifacts; harmonics put 150 Hz-1 kHz at -15 dB share (survives phones) pro_sub_drop bench not independent (targets from that trim)
hover_tick good inside 6/7 (real 5), tone 21.0 (real x 21.2), var 0.807 (forge 0.799) critic 0/5 (totals -0.18..1.31 vs refs 0.78..2.44)
glitch ok (no refs) critic 3/4, var 0.799, decorrelated, 0 artifacts 04 hit weak; only 4 finals
click ok t40 x 1.22, env 0.24, var 0.681 (v1 0.886) narrow partial lines (6 kHz tink, 250-500 Hz body) ring for the whole 80-130 ms in 3 of 5; 1-4 kHz share -14 vs bench -4.9 dB; critic 1/5
pop ok inside 7/7, var 0.735, env 0.55 no bright click layer: 1-4 kHz share -17.2 vs bench -2.0 dB; tone and AudioSet worse than v1 (11.4 vs 8.6 dB, 0.4 vs 1.0)
notify ok inside 7/7 (real 5.5), AudioSet "Ding" top-1 5/5, curve 21.8 (real x 28.9) pitched 1-2 oct above the bench bells (1-4 kHz share -0.1 vs 150 Hz-1 kHz in the bench), no strike transient, t40 x 0.63; critic 0/5
pro_notify ok balance overlays the 4 film dings (2 kHz + 3.4x partial), AudioSet "Ding" 5/5 critic -5.7..-7.2 vs films 0.1..3.0 (CLAP "cheap beep"); var 0.882
toggle ok inside 5/7, tone 8.82, var 0.72 150 Hz-1 kHz share -5.4 vs bench -19.2 dB (14 dB too much body); second contact is the peak (env oct 2.58)
impact ok tone 6.29 (real 6.19), t40 x 0.62 (v1 0.50), AudioSet "Explosion" 4/5 dull onset: >6 kHz share of the first 15 ms -28.9 vs bench -21.6 dB; 1-4 kHz 8 dB thin; critic 1/5
pro_impact ok critic 3/5, granular sub body matches the films 15 dB less grain sparkle (4-12 kHz share -30.4 vs films -15.4); the body decays too little before the gate (last 50 ms at -14..-25 dB re peak vs the films' -29..-30)
riser ok inside 3/6 (real 4), var 0.78 (bench 0.767), AudioSet 0.8, critic 5/5 riser_05 has a steady 2.5 kHz line for 2.5 s; riser_02 a steady sub drone line
pro_riser ok- var 0.676, critic 4/5 (maker) inside 3/6 (real 6); AudioSet "Music" 4/5, CLAP "retro game": reads as a chiptune arpeggio
shimmer ok- inside 6/7, tone 10.5 (real x 9.7) tails half the bench (t40 1.16 vs 2.42 s); steady narrow partial lines; chimes corr 0.97
pro_shimmer ok- var 0.646, sparkle 01 looks like the film runs flagged (fold, resonance); critic 1/5
pro_pop ok- inside 4/6 (real 3), stepped chirp matches the film critic 0/5 (best 1.57 vs film refs 1.56 / 0.19); CLAP "chiptune"
pro_click ok- critic 5/5 vs its one forge ref vs the two held-out film clicks: 150 Hz-1 kHz share -41 vs -6 dB: bright ticks without body (tone 25.8 vs bench loo 18.5); var 0.870
pro_tap ok- study targets met (maker) critic 0/5, one ref, no bench
pro_logo_sting ok- chirp-to-hum timing matches the film sting critic 0/4; inside 3/6 vs the bench end chord (a different sound)
pro_hover_tick weak inside 4/7 (real 6), tone 7.9 (real x 6.5) phrase z 0.85; critic 0/5; var 0.878
pro_glitch weak critic 3/5, balance matches the film too even and dense: median 2 ms level -19..-24 dB re peak vs the film's -31 (less burst-to-burst drama); 01 adds a 50-200 Hz block; var 0.875
tap weak tone 6.1 (real x 9.35), t40 x 1.27 phrase z 0.52 (CLAP does not hear a fingertip on glass), inside 3/7 (real 6), var 0.856 (bench 0.716); critic 0/5
swish worst 3 (with whoosh_fast) inside 4/7 (real 6), t40 x 0.68 4-12 kHz share -18.2 vs bench -33.5 dB (15 dB hiss); AudioSet top-1 "Whack, thwack" 3/5 and "Knock" 0.76: an impact-like onset; critic 0/5
whoosh_fast worst 3 inside 4/7 (v1 7/7), typ z -4.85 tail dies 2x too fast (t40 130 vs 265 ms, x 0.49, v1 0.51); <150 Hz share -9.7 vs -17.7 dB (8 dB boom); 4-12 kHz -17.4 vs -32.3 dB (15 dB hiss); AudioSet "Knock" 2/5; critic 0/5
logo_sting worst 3 inside 3.5/6 (real 5.5), tone 16.3 (real x 15.5) rings 2.5x too long (t40 2.14 vs 0.86 s); bell chords where single-ring refs have one partial set; critic 0/4; the refset (struck rings) is not a composed launch sting
swipe worst inside 4/7 (real 6), tone 13.5 vs bench loo 3.6 muffled: 4-12 kHz share -35.9 vs bench cloth -19.2 dB, onset >6 kHz -27 vs -11.5 dB; AudioSet "Knock" / "Door" on 2 of 5 (reads as a thump); critic 0/5; the forge refs are just as dark (-39 dB), so the forge split steered it away from the bench

Best three: whoosh_big (+ pro_whoosh_big), typing_key (+ pro_typing_key / run), pro_pop_in. Worst three: swipe, whoosh_fast + swish (one engine, one fault), logo_sting.

Integration changes in this round (judge)

  • bench.py additions (no change to the 18 existing rows): --made reads <category>/final/ when present and skips audition*.wav / _*.wav (four areas' first bench runs had counted audition.wav as a made sound); phrasings, confusions and AudioSet labels for the 16 pro_* categories; pro_riser ends on its peak like riser.
  • After the first full run, notify_04.wav was re-rendered from its report params with the current recipe (identical after 10 ms; the fade-in removes the start click the bench flagged). logo_sting / pro_logo_sting audition.wav rebuilt without the rejected 05 variants. The table above is the second full run, after that fix (24 s with cached embeddings; the first took 375 s). Remaining artifact flags: 3, all by design: impact_01 and logo_sting_01 "late start" (a reverse swell / pre-swell before the hit: place them by the hit, not the file start) and pro_typing_run_04 "tail cut" (the run's low swell ends on the cut, as in the film).