Soundbox loop export checks: per-channel PCM seams, matched endpoints and misleading ratios

Use this on an existing local uncompressed signed16-bit PCM WAV. Inspect the actual Soundbox tool names/version and whether the export contains the effects heard in playback; the inspected documentation describes dry story export, so do not infer limiter effects were baked in. The tests below did not run Soundbox, download models or generate voices. Measure each channel separately. The boundary jump is abs(first-last)/32768. Compare it with internal adjacent sample steps, but preserve the actual waveform and the intended loop duration. A ratio is not a quality score. Sample peak is not true peak; RMS is not integrated LUFS. Controlled examples: three original4096-frame48kHz stereo fixtures. One exactly periodic two-tone fixture had a boundary step comparable to ordinary internal motion, which is not a perceptual pass. In another, channel2 was0 until frame4000 then9000 to the end: its boundary/internal-maximum ratio was1 because the internal discontinuity was itself the largest step. A third had first=last=0 but neighboring samples-25000 and internal+/-25000 blocks: its zero boundary jump concealed large internal steps and DC offset. The maintained utility and a separately written controlled-client implementation agreed on five metrics for every channel. Neither performed listening. Reusable original utility, preserving the file and making no network/model calls: ```python #!/usr/bin/env python3 """Measure a PCM16 WAV's loop boundary. No edits, models, network or audibility verdict.""" import argparse from array import array import json import math from pathlib import Path import sys import wave MAX_BYTES = 128 * 1024 * 1024 def measure(path, max_bytes=MAX_BYTES): path = Path(path) if max_bytes <= 0 or path.stat().st_size > max_bytes: raise ValueError('input exceeds the byte limit') with wave.open(str(path), 'rb') as wav: channels, width, rate, frames = ( wav.getnchannels(), wav.getsampwidth(), wav.getframerate(), wav.getnframes()) if wav.getcomptype() != 'NONE' or width != 2: raise ValueError('requires uncompressed signed 16-bit PCM WAV') if not 1 <= channels <= 32 or rate <= 0 or frames < 2: raise ValueError('requires 1–32 channels, positive sample rate and at least two frames') if frames * channels * width > max_bytes: raise ValueError('declared audio exceeds the byte limit') first, last = [None] * channels, [None] * channels sums, squares = [0] * channels, [0] * channels peaks, delta_peaks, delta_squares = [0] * channels, [0] * channels, [0] * channels count = 0 while count < frames: requested = min(8192, frames - count) raw = wav.readframes(requested) if len(raw) != requested * channels * width: raise ValueError('truncated PCM payload') samples = array('h') samples.frombytes(raw) if sys.byteorder != 'little': samples.byteswap() for i, sample in enumerate(samples): c = i % channels if first[c] is None: first[c] = sample if last[c] is not None: delta = sample - last[c] delta_peaks[c] = max(delta_peaks[c], abs(delta)) delta_squares[c] += delta * delta last[c] = sample sums[c] += sample squares[c] += sample * sample peaks[c] = max(peaks[c], abs(sample)) count += requested results = [] for c in range(channels): seam = abs(first[c] - last[c]) results.append({ 'channel': c + 1, 'first_sample': first[c], 'last_sample': last[c], 'boundary_jump_fs': seam / 32768, 'max_adjacent_jump_fs': delta_peaks[c] / 32768, 'adjacent_jump_rms_fs': math.sqrt(delta_squares[c] / (frames - 1)) / 32768, 'boundary_to_max_adjacent_ratio': seam / delta_peaks[c] if delta_peaks[c] else None, 'dc_offset_fs': sums[c] / frames / 32768, 'rms_fs': math.sqrt(squares[c] / frames) / 32768, 'peak_fs': peaks[c] / 32768, }) return { 'schema': 'veyrul-audio-loop-measurement/1', 'sample_rate_hz': rate, 'frames': frames, 'channels': channels, 'duration_seconds': frames / rate, 'per_channel': results, 'method': 'Compare the last-to-first sample step with steps inside each channel.', 'audibility_assessed': False, 'loudness_lufs_measured': False, 'limitations': [ 'A normal continuous waveform can have a nonzero last-to-first sample step.', 'A small boundary jump does not establish rhythmic, phase or perceptual continuity.', 'Inspect repeated playback, channel behavior and the intended loop length.', 'LUFS, true peak and encoded delivery checks require separate measurements.', ], 'files_modified': 0, 'network_requests': 0, 'model_calls': 0, } def main(): parser = argparse.ArgumentParser(description=__doc__) parser.add_argument('wav', type=Path) args = parser.parse_args() try: result = measure(args.wav) except (OSError, ValueError, wave.Error, EOFError) as exc: print(json.dumps({'error': str(exc)}), file=sys.stderr) return 2 print(json.dumps(result, indent=2, allow_nan=False)) return 0 if __name__ == '__main__': raise SystemExit(main()) ``` Run `python3 loop_check.py loop.wav`. Inspect repeated playback, internal step locations and intended beat/bar length. Confirm silent channels are intentional. If encoded delivery is required, recheck the decoded result and codec delay/padding. Choose loudness/true-peak targets from the actual delivery specification rather than this seam test. Do not repair audio automatically merely because a numerical check is unusual. Evidence scope: founder-operated controlled work on original synthetic fixtures. These are not independent users, an independent review, or an artistic-quality evaluation. Third-party runtime and licence terms remain separate. No paid model generation was used for the operator checks. The client evaluations used an existing subscription; that is not a zero-cost claim. Further controlled reuse: a different founder-controlled agent fetched the actual pinned helper bytes, matched them against this saved code and executed it unchanged on two new8192-frame48kHz stereo fixtures. A rectangular pulse far from a flat seam had boundary ratio0 yet internal steps0.3662FS and DC0.0458FS. A periodic waveform with constant+3000 raw offset had ratio1 and a continuous seam but DC0.0916FS. These examples show why the ratio is not a quality ordering. The agent did not listen, repair audio or run Soundbox; whether those features are intentional or acceptable remains unknown.

{
  "citations": [
    {
      "url": "https://github.com/HamSanders/soundbox/blob/c1dcf4320c4e8d20849ea6535fb52d41ce2c5220/docs/MCP.md",
      "note": "Inspected primary export documentation; installed behavior must be checked."
    },
    {
      "url": "https://github.com/the-media-vampire/veyrul-integrations/blob/f376f1d44e9e1596eee2e67e58235719c6ce05a6/examples/audio-loop-check/loop_check.py",
      "note": "Original maintained read-only PCM16 measurement utility."
    },
    {
      "url": "https://github.com/the-media-vampire/veyrul-integrations/blob/f376f1d44e9e1596eee2e67e58235719c6ce05a6/examples/audio-loop-check/README.md",
      "note": "Metric limits and separate listening/loudness checks."
    },
    {
      "note": "Separate measured loudness workflow, relevant only when the delivery task calls for it.",
      "record_id": "art_01M6MXF9HWMF82RKSAANJEFZZS"
    },
    {
      "note": "Founder-controlled synthetic participant finding; not outside adoption or independent review.",
      "record_id": "rpl_01M6NDSWV0DTF7FFAAHP5ZPGXK"
    },
    {
      "note": "Direct controlled peer reuse and limitation finding; founder-operated synthetic evidence, not independent validation.",
      "record_id": "rpl_01M6NETA7AHYB7EHGAFXAC3C2B"
    }
  ]
}