Your China-based production partner sends a smoother revision of an appliance comparison. The loudness no longer jumps between products. That may improve the edit, but does it still answer which product was quieter in the recording?
ONCE recommends specifying the listening question before commissioning the mix. Preserving relative recorded levels and discussing sound character at matched loudness are different tasks. A single unlabeled audio version should not silently switch between them.
Matching levels is what the tool is meant to do
Adobe’s Auto-match audio loudness guide, updated January 7, 2026, describes consistent volume across clips and displays clip loudness in LUFS. That is a mixing function, not an appliance-noise measurement method.
The Audio Engineering Society’s Loudness Normalization resource distinguishes individual track normalization from applying the same gain across a group to retain internal level relationships. Its context is audio distribution. Applying that distinction to product-video handover is ONCE’s recommendation, not an AES appliance-test standard.
A brief asking only for balanced, polished audio may invite an editor to reduce the very difference the buyer wants to hear. Make preservation of the paired recordings’ relative levels explicit.
Two listening versions answer different questions
A relative-level version preserves the relationship in recordings made under stated, comparable conditions. If an overall playback adjustment is needed, apply one fixed gain to the group, avoiding clipping and additional dynamics processing. Do not normalize each sample to the same target. This retains the recording relationship; it does not establish universal acoustic superiority.
A loudness-matched version supports a different discussion: how the sounds are characterized after their levels have been adjusted. A team might identify a periodic rattle, a high whine or steady airflow. Label the matching and retain each adjustment. Do not use that version’s amplitude, waveform or preference vote to reconstruct the original noise ranking. Matching alone does not control every perceptual or playback variable.
Neither version converts a player setting, headphone impression or LUFS reading directly into an on-site sound-pressure level. Numerical acoustic claims and broader superiority claims need their corresponding approved measurement material and conditions.
ONCE’s listening-purpose sheet comes before the mix
Record the question, comparable product states, selected time ranges, source files, permitted processing and intended advertising conclusion. Have the sound editor and brand reviewer agree before exporting. Buy both versions only when both questions matter.
Imagine a fictional air-purifier comparison. The product team defines comparable operating conditions; the recorder maintains the stated microphone position and settings. A different setting on one machine cannot be made equivalent merely by changing audio levels. Preserve a relative-level version first. If the team separately wants to discuss tonal character, make a clearly identified matched version from those same sources.
Keep a short processing record: gain per sample, noise reduction or compression, selected operating interval and any replaced source. Identical fader positions do not restore a relationship lost during recording. Where the original conditions or automatic level behavior cannot be established, narrow the comparison rather than claiming reconstruction.
Retain the originals. Music, narration, generated effects and creative enhancement can belong in the advertisement, but should not overwrite the listening master or silently replace the product sound under comparison.
Approve the question before approving the polish
For a quieter-product claim, inspect separate normalization and asymmetric processing. For sound-character discussion, inspect how matching and listening conditions are described. Then review the actual publishing context. Two independently processed uploads should not automatically be treated as the original paired comparison.
Scope paired capture, the purpose sheet, processing records, each needed listening version and the advertising mix as identifiable work. “Make both sound good” is not acceptance of a comparison task. A production recording also should not be sold as a new acoustic test. Find related planning in ONCE’s video production articles.
The reusable asset for managed content services is the connection between original sound, processed versions and approved purpose. It helps the next editor know which file supports which decision. Any reduction in rework needs actual project evidence.
Scope: Official materials checked November 8, 2026. The AES page displays no publication date. The listening method and appliance example are ONCE recommendations and hypothetical planning, not client results, product tests or certification. The original AI cover is conceptual.
