Seams, tags and session boundaries

调调 gain-tag checks: distinguish ReplayGain metadata from a sample-altered copy

Prepare an owned FLAC original, a metadata-only working copy and another with externally reduced sample amplitudes. Inspect processing evidence before observing identifiable playback in 调调. A tag does not guarantee player application, and similar sound does not establish identical file bytes.

English translation of the Chinese original, translated on . Original publication dates, attribution, code and version notes are retained.

Prepare an owned FLAC original, a metadata-only working copy and another with externally reduced sample amplitudes. Inspect processing evidence before observing identifiable playback in 调调. A tag does not guarantee player application, and similar sound does not establish identical file bytes.

Preserve one original and two processing branches

Choose an owned short recording with a decodable original FLAC and retain source/project. Create separate tag-copy and half-copy branches: metadata-only versus changed sample amplitudes, without combining operations. Use an identifiable opening and moderate peaks, requiring no high-volume/full-scale test. Keep original title and external branch label separate. Names are preparation plans, not supplied downloads. Without suitable tools preserve the original and mark each unavailable processing branch pending.

Identify ReplayGain as gain metadata

Xiph’s metaflac documentation describes calculating track/album gains and peaks and storing them as FLAC tags. Do not invent scan results. If an existing tool is used, follow its actual instructions on tag-copy only and save real output, noting replacement of related existing tags. Keep originals outside the operation. Without that tool, inspect an authorized documented working copy or leave fields unknown. Scan result, tag text and player application are separate states. Listed FLAC import does not promise every ReplayGain interpretation in 调调.

Check whether the tag branch preserves decoded samples

Compare original/tag-copy decoded audio externally: rate, channels, frame count and full PCM when suitable tools exist. Keep the decoded output format fixed for the sample comparison. Whole-file hashes may differ after metadata changes without implying changed decoded samples. Unchanged size alone proves no audio identity. Claim sample equality only from an executed reliable comparison; otherwise retain metadata-operation scope and leave samples unverified. A FLAC internal checksum is no universal edited-file security certificate. This branch separates metadata from audio changes rather than predetermine player volume.

Halve samples in the separate branch under an explicit model

Actually reduce amplitudes externally in half-copy. The paper model stipulates y=0.5x, e.g. 0.6 → 0.3, −0.4 → −0.2 and 0 → 0. If a tool accepts decibels, follow its conversion and retain the actual value; integer −6 dB is not an exact mathematical synonym for halving. Fix rate/channels/interval and avoid unverified gain tags confounding this branch. Integer formats introduce rounding; sample examples promise no byte-wise halving. Save processing/export evidence rather than call a lowered player slider changed source samples.

PAPER SAMPLE MODEL ONLY: y = 0.5*x
x:  0.6   -0.4   0
y:  0.3   -0.2   0
metadata-only branch: inspect actual tags and decoded PCM separately
baked-gain branch: inspect actual changed samples and export conditions
No player ReplayGain application is presumed.

Separate import deduplication from gain-tag support

Import readable working copies through actual controls, recording added/deduplicated/failed outcomes. A tag-only copy may encounter content deduplication, whose algorithm is undisclosed; this cannot identify hash-versus-decoded-audio comparison. Without a distinguishable tag-copy entry, leave that player comparison unestablished instead of substituting the original. Verify half-copy identity as well. Do not repeatedly transcode, rename or replace the recording to force three rows, which destroys the controlled common source.

Fix output and volume while recording limited observations

Only for distinguishable sources, play the same interval at normal speed through fixed output and the same comfortable volume setting. Verify title/anchor before recording perceived differences; do not adjust each file to equal loudness. Similar sound in tag-copy does not prove that the tag was ignored, and quieter sound alone does not prove that ReplayGain was applied; other processing may intervene. Reduced external amplitudes do not guarantee perceptually half loudness. Without a loudness meter claim no LUFS, peak or sound-pressure measurement. Inconclusive perception is a valid scoped observation.

Diagnose branch and identity confusion

For three apparently identical results inspect entry identity, deduplicated tag-copy and whether half-copy merely changed a name or exported an unprocessed interval. Large differences require checking extra normalization, tags or wrong source portions externally, not endless system-volume adjustment. Inspect one branch/factor while retaining originals and initial responses. Unknown tag fields remain source text/unknowns rather than a fabricated professional-looking set. External sample equality certifies those files only, not bit-perfect phone output or automatic loudness normalization across every service.

Deliver three evidence layers and unpromised-feature limits

Deliver metadata processing/fields, external decoded-sample comparison and identifiable player observations as three layers, noting model-versus-real-tool differences in half-copy. Explain distinct whole-file bytes, audio samples and perceived loudness, marking unestablished branches pending. Restore ordinary level and retain originals/project. Invent no unannounced ReplayGain switch, built-in scanner/editor or loudness certification. New tools, tags or outputs require renewed checks rather than a universal gain-metadata support claim.

Decoded lossless samples versus playback; Same-content import outcomes.

Checked on 19 November 2026: Apple’s Russia lookup and China lookup both list 调调 1.0.2, requiring iOS / iPadOS 17.6 or later. The official store describes audio import, playback, lyrics and music themes and Home Screen widgets. Use owned or authorized material and controls actually visible in the installed version. Keep external preparation, paper expectations and real observations separate; no completed device, private-server or exhaustive format-compatibility test is claimed.

Method references: Xiph metaflac metadata and ReplayGain; Audacity external amplification.