Measure program loudness and true peak separately. Use the delivery specification or a clearly stated house target; do not assume that a popular LUFS number is a universal requirement for YouTube, podcasts, social video and broadcast.
On this page
Read the meters correctly
| Measurement | Question it helps answer |
|---|---|
| Integrated loudness | How loud is the measured program overall? |
| Short-term or momentary loudness | How does loudness change within the program? |
| Sample peak | What is the largest stored sample amplitude? |
| True peak | What peaks does a reconstructed-signal estimate identify? |
The integration window matters. Measuring one loud paragraph is not the same as measuring an entire episode with pauses, music and a quiet ending. Reset and run the measurement over the complete delivery range when the target refers to a full program.
A standard is not a platform-wide rule
EBU R 128 uses a −23 LUFS program-loudness recommendation in its broadcast context. That does not make −23 LUFS the required target for every creator upload. Read the actual recipient’s specification, including peak and tolerance requirements, before adapting a mix.
Build a repeatable delivery check
- Write down the target, meter mode and reason for choosing them.
- Balance dialogue and music by listening before trying to hit one integrated number.
- Measure the complete program and inspect unusually loud or quiet sections.
- Adjust gain and dynamics only as needed; note when peak control changes the waveform.
- Export the chosen delivery codec, reopen the file and measure again.
- Listen to the result at a normal playback level on the intended listening setup.
Why a target alone cannot guarantee a good mix
Two programs can share integrated loudness while having very different dynamics, background noise or dialogue intelligibility. Turning a noisy recording up can make the problem more obvious. Excessive limiting can meet a number while making the result tiring. Use the meter to support an audible decision rather than replace it.
The first useful fix may be clip-level balancing, a repaired transient or better source recording. See normalization methods for the distinction between a gain change and wider processing.
Sources and scope
This is an explanation of measurement and delivery methodology. It makes no claim that Creator Edit Lab measured platform normalization behavior. Sources checked October 6, 2026.