LUFS, dBFS and True Peak: How to Measure a Voiceover Delivery
Updated: Nov 5
Use LUFS to describe programme loudness and peak measurements to check different aspects of signal level. A voiceover can meet a sample-peak ceiling and still sound too quiet, or reach a loudness target while exceeding a true-peak limit. Start with the destination's written requirements, measure the intended delivery and listen; there is no single loudness number appropriate for every project.
Separate the three measurements
dBFS expresses digital level relative to full scale. A sample-peak meter reports the largest stored sample level within its measurement interval. Integrated LUFS describes loudness across a programme using a defined measurement model, rather than simply reporting the tallest sample.
True peak estimates the reconstructed waveform's peak between samples and is commonly expressed in dBTP. It is different from a sample-peak reading. Use a meter that explicitly supports the measurement required by the destination instead of assuming a generic green/red meter provides every number.
Do not confuse a negative target with an instruction to remove clarity. A file measured at −23 LUFS is not automatically a worse recording than one at −16 LUFS. Those values can reflect different delivery needs and different gain, dynamics or content.
Get the receiving specification first
Ask whether the requirement applies to a standalone voice stem, a complete programme or a finished video mix. Confirm integrated loudness, permitted true peak, channel layout, file format and any tolerance or short-term requirement. A screenshot saying “audio should be loud” is not an actionable technical specification.
For an example of a defined standard, EBU R 128's short-form supplement specifies programme loudness of −23 LUFS and a maximum true-peak level of −1 dBTP for linear audio, with additional conditions. These broadcast values are an example of why specifications matter, not a universal preset for social media narration.
If a client provides no loudness target, agree on the intended playback and delivery requirements rather than inventing a mandatory platform rule. Avoid treating a number quoted by an unrelated tutorial as a contract for your project.
Measure the right programme from a reset meter
Place the loudness meter at the relevant output stage and reset its measurement before playing the full intended delivery. Include the programme boundaries specified for the job. A ten-second excerpt and the complete narration can yield different integrated readings.
For a finished video with music, measure the final mix as required, not only the voice soloed on its track. For a requested voice stem, measure that stem separately. Write down which file and channel layout produced each reading so later comparisons remain meaningful.
Check mono and stereo handling in the meter and export. A routing change or duplicated channel can affect a measurement. Do not compare two numbers as if they describe the same signal when one is a mono stem and the other a stereo programme.
Write the delivery limits beside the meter readings
Make three fields for integrated loudness, sample peak and true peak. Measure the complete approved file, then compare each field with the corresponding delivery requirement. If only one reading misses its limit, adjust for that reading and measure the export again.
Calculate a provisional level change
If an illustrative file measures −22 LUFS and the agreed target is −18 LUFS, the loudness difference is 4 LU. A simple 4 dB gain increase is a starting calculation when dynamics remain unchanged, not an automatic safe correction. If its measured true peak is −2 dBTP, that gain would predict approximately +2 dBTP before other processing, conflicting with a negative true-peak ceiling.
This example shows why the peak and loudness checks must be considered together. It does not prescribe compression or limiting settings. A mix revision, controlled dynamics or another delivery decision may be needed, followed by new measurements and listening.
Audacity documents Loudness Normalization separately from ordinary peak normalization. Choose the effect that matches the task and verify the result; a tool that sets one metric does not establish all delivery criteria by its name alone.
Measure and listen after export
Export the required format from the saved project and inspect that actual file. Encoding can affect peaks and other properties, so do not assume the timeline reading is the final delivery result. Recheck programme loudness and true peak with suitable tools, then listen on normal playback equipment.
Keep a short delivery record with filename, specification, measured integrated loudness, peak type and value, channel layout and review date. Do not invent measurement results when a meter has not been run.
AI LEMNISCATA currently supplies MP3 narration without a promised LUFS target or mastering control. Prepare and create the voice there when appropriate, then perform any required delivery measurement and mastering in a suitable editor. Review the current order total before generation.
Common questions
Does normalizing peaks to −1 dBFS set LUFS too?
No. Peak normalization changes level according to peaks. The resulting integrated loudness depends on the content and its dynamics.
Are LUFS and playback volume the same thing?
No. LUFS is a file/programme measurement; a listener's device volume and playback environment still affect what they hear.
Should every narration match a published streaming target?
Only when that target is actually applicable to the destination and deliverable. Follow current receiving requirements, not a universal number copied across projects.
Related guides
Sources and scope
Sources checked 22 October 2026. The diagnostic workflow and hypothetical examples are independently written; no paid generation or claimed repair test was performed for this guide.
EBU: R 128 short-form loudness specification
Audacity: Loudness Normalization
Create on AI LEMNISCATA
Use the current workspace to prepare your next voiceover script. Check the available choices and displayed total before ordering. Any external editing or technical delivery checks described above remain separate.