Why Can an Encoded Product Ad Exceed My Audio Peak Limit When the Editing Master Did Not?

Identify which peak was measured

First record the meter type used for the editing master. A sample-peak meter reports the largest stored sample values. A true-peak meter estimates peaks that may occur between samples when the signal is reconstructed.

ITU-R BS.1770 explicitly distinguishes these measurements and describes oversampling-based true-peak estimation. A signal can have samples below a specified ceiling while its reconstructed waveform reaches a higher level.

The difference is especially relevant when a delivery contract specifies a true-peak ceiling in dBTP. A sample reading in dBFS is not automatically equivalent to a true-peak reading.

Do not treat an ordinary loudness reading as a peak measurement. Integrated loudness, sample peak and true peak answer different questions.

Compare the master and encoded file

Preserve the approved uncompressed or high-quality editing master. Record its sample-peak and true-peak readings using a clearly identified measurement method.

Encode the intended delivery copy with the required audio codec and settings. Decode or analyze that actual file with a true-peak-capable meter. Record the result separately from the master.

If the master meets the specified ceiling but the encoded copy exceeds it, the final encoding stage requires investigation. Do not assume that changing the filename or lowering the playback volume fixes the delivered signal.

FFmpeg's loudnorm filter documents a maximum true-peak target and true-peak detection. Its normalization behavior depends on the selected mode and other conditions, so a requested setting is not proof of the final measurement.

Hypothetical narration-and-music delivery

Hypothetical illustration—not an executed encoding test: a product advertisement's editing master has a sample peak of -1.5 dBFS. The recipient specifies a true-peak ceiling of -1.0 dBTP.

The sample-peak figure alone cannot establish compliance with that requirement. The editor therefore measures the master's true peak and the encoded delivery's true peak separately.

Suppose the fictional encoded copy is reported at -0.6 dBTP. Under the stated hypothetical ceiling, it exceeds the permitted value by 0.4 dB. The next step is to revise the finishing or encoding approach and measure the new delivery file.

These numbers are invented inputs, not results from a real advertisement or universal publishing limits.

Keep the ceiling tied to the receiving specification

Use the actual recipient's requirement, including its measurement method and any specified tolerance. Do not prescribe one true-peak ceiling for every platform.

After adjusting the finishing chain, measure the complete final file again. Listen for limiter artifacts, changed music transients and speech distortion. A numerically acceptable peak does not guarantee that the advertisement sounds natural.

Record the codec, file identity, sample rate, meter and final result. If the destination processes the upload again, inspect its available output separately rather than assuming the production-file measurement survives unchanged.

The existing loudness article distinguishes peaks from whole-program loudness. This guide isolates the additional boundary between stored sample peaks, reconstructed true peaks and the encoded deliverable.

Record your own observations

The filled rows are illustrative examples. Use the blank row for your own files and observations.

Download the editable CSV worksheet

What the references establish

• ITU-R BS.1770-5 — Algorithms to measure audio programme loudness and true-peak audio level: Primary standard defining true-peak measurement and distinguishing it from sample-peak measurement.
• FFmpeg Filters Documentation — loudnorm: Documents true-peak targeting, measurement and normalization-mode behavior.