62d6732504
Two root causes, both found by building an oracle rather than guessing. 1. Amplitude ~75x low: the IMDCT applied a 2/n normalisation the encoder's forward transform had already carried. Because 2/n varies with block size it produced DIFFERENT errors on 256- vs 2048-sample blocks — exactly the reported symptom. Removing it gives fit scale 1.0000. 2. Leading trim, the real remaining defect. The first audio packet produces NO output (its window only primes the overlap-add) but we emitted from the first block's centre, injecting half a priming window of garbage and shifting everything early. And Vorbis carries encoder delay in the GRANULE POSITION, which varies per file — afinfo confirms 128 / 1103 / 960 frames on three samples — while our Ogg reader kept only last_granule and discarded per-page granules, making it unrecoverable. Added per-page granule tracking: the first page reporting a granule pins priming as centre - granule, and valid audio starts at priming + blocksize_0/2. That reproduces afinfo's numbers exactly on all three. A premise in the brief was also wrong and worth recording: our output length was already correct. afinfo reports valid frames matching OUR output — it is afconvert that trims a further 128. The reference WAV was short, not us. mono 47 files mean corr 1.00000 (min 1.00000) 47/47 exact stereo 107 files mean corr 0.826 68/107 exact Decode cost 5.13 ms/file average; 11.5 MB compressed expands to 143.3 MB of f32 PCM, which is why the sample bank's LRU cap matters. Honest remaining defect: ~39 stereo files decode wrongly and it is NOT alignment — a full lag sweep peaks at 0.40-0.89 with fit scales 0.40-1.87, so specific blocks have wrong amplitude. Mono being 47/47 rules out floor, residue 0/1, MDCT, windowing and priming; coupling matches the spec's square-polar mapping including reverse order; floor 0 is rejected rather than mis-decoded; and both channels are identical in the failing files, so it is not a swap. The failing set is transient-heavy impact/footstep sounds, so the lead is residue type 2 partition counting on short blocks. reference_decode.rs is no longer #[ignore]d: 3 real tests asserting mono correlation > 0.999 and length == granule, plus a 3000-mutation fuzz that must never panic, all skipping cleanly without fixtures. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
86 lines
2.9 KiB
Rust
86 lines
2.9 KiB
Rust
//! Derive the trim rule: search (start, len) against a reference so the
|
|
//! correct offset is measured, not assumed.
|
|
//! Usage: vorbis_trim <file.ogg> <ref.wav>
|
|
use makepad_game_audio as audio;
|
|
|
|
fn main() {
|
|
let a: Vec<String> = std::env::args().collect();
|
|
let ogg = std::fs::read(&a[1]).expect("ogg");
|
|
let refw = std::fs::read(&a[2]).expect("ref");
|
|
let (buf, first_center, granule) = audio::vorbis::debug_raw(&ogg).expect("raw");
|
|
let want = audio::wav::decode(&refw).expect("ref wav");
|
|
let ch = buf.len();
|
|
let buflen = buf[0].len();
|
|
let wf = want.frames();
|
|
println!("buflen={buflen} first_center={first_center} granule={granule} ref_frames={wf} ch={ch}");
|
|
|
|
// Exact-match search: for each candidate start, how well does
|
|
// buf[start .. start+ref_frames] line up with the reference?
|
|
let mut best = (0usize, f64::INFINITY);
|
|
let lo = first_center.saturating_sub(1024);
|
|
let hi = (first_center + 1024).min(buflen.saturating_sub(wf));
|
|
for start in lo..=hi {
|
|
if start + wf > buflen {
|
|
break;
|
|
}
|
|
let mut err = 0f64;
|
|
// Sparse but dense enough to rank candidates unambiguously.
|
|
let mut i = 0usize;
|
|
while i < wf {
|
|
for c in 0..ch {
|
|
let g = buf[c][start + i] as f64;
|
|
let w = want.samples[i * ch + c] as f64;
|
|
err += (g - w) * (g - w);
|
|
}
|
|
i += 7;
|
|
}
|
|
if err < best.1 {
|
|
best = (start, err);
|
|
}
|
|
}
|
|
let rms = (best.1 / (wf / 7).max(1) as f64).sqrt();
|
|
println!(
|
|
"BEST start={} (first_center{:+}) rms_err={:.3e} | granule-reflen={}",
|
|
best.0,
|
|
best.0 as i64 - first_center as i64,
|
|
rms,
|
|
granule as i64 - wf as i64
|
|
);
|
|
println!(
|
|
" tail: buflen-(start+reflen)={}",
|
|
buflen as i64 - (best.0 + wf) as i64
|
|
);
|
|
// Is the region our decode emits before the reference's first sample
|
|
// actually silent? If so afconvert trimmed it and we are not wrong.
|
|
let rms_of = |lo: usize, hi: usize| -> f64 {
|
|
if hi <= lo {
|
|
return 0.0;
|
|
}
|
|
let mut s = 0f64;
|
|
let mut n = 0usize;
|
|
for c in 0..ch {
|
|
for i in lo..hi.min(buflen) {
|
|
s += (buf[c][i] as f64).powi(2);
|
|
n += 1;
|
|
}
|
|
}
|
|
if n == 0 { 0.0 } else { (s / n as f64).sqrt() }
|
|
};
|
|
let peak_of = |lo: usize, hi: usize| -> f64 {
|
|
let mut p = 0f64;
|
|
for c in 0..ch {
|
|
for i in lo..hi.min(buflen) {
|
|
p = p.max((buf[c][i] as f64).abs());
|
|
}
|
|
}
|
|
p
|
|
};
|
|
println!(
|
|
" lead region [{}..{}]: rms={:.3e} peak={:.3e} (signal rms={:.3e})",
|
|
first_center,
|
|
best.0,
|
|
rms_of(first_center, best.0),
|
|
peak_of(first_center, best.0),
|
|
rms_of(best.0, best.0 + wf.min(4096))
|
|
);
|
|
}
|