Measure Loudness the Way Platforms Measure It
Drop an audio file and you get its integrated loudness in LUFS, its loudness range, its sample peak, and how far it sits from the targets Spotify, YouTube, Apple Podcasts and broadcast use.
The measurement runs on your device. Nothing is uploaded, which matters when the file is an unreleased episode or a master you are still working on.
LUFS Is Not the Same as Volume
This is the distinction the whole tool rests on, and it is the reason a level meter in an editor will not answer this question.
A plain level meter measures signal. It tells you how big the numbers in the file are. LUFS measures how loud the audio seems to a listener, which is a different question with a different answer. Two files can have identical peaks and identical average levels and still sound unlike each other in volume.
The standard closes that gap in two steps. First it filters the audio to approximate how human hearing weights frequencies, treating a boomy low end as less loud than the midrange where speech lives. Then it discards the quiet parts before averaging, which is the part that catches people out.
Why Silence Would Otherwise Ruin the Number
Take an interview with a long pause in the middle. Average the whole file and the pause drags the average down, so the file measures quiet even though every word in it was at a sensible level. Add a longer pause and it measures quieter still. The audio has not changed at all.
The standard fixes this with gating. Blocks below an absolute floor are thrown out, then a second pass throws out anything far below the level of what remains, and only the surviving blocks are averaged. The result is the loudness of the content rather than the loudness of the content diluted by its gaps.
That is the single biggest practical reason to use a LUFS reading instead of an average level. A pause, a cold open, or a quiet outro no longer skews the answer.
The Targets, and What Happens If You Miss Them
Streaming services normalize playback to a target so listeners are not adjusting the volume between tracks. The figures they publish sit around minus fourteen for music streaming and video, a little lower for spoken podcast audio, and lower again for broadcast.
Being louder than the target is not rejected, which is why it goes unnoticed. The platform simply turns your upload down. Everything you did to make the master feel loud gets undone at playback, and what survives is a flatter, more compressed version of it. Mastering hot no longer wins anything; it only costs you dynamics.
Being quieter is the opposite problem. Some services raise quiet audio and some leave it alone, so a quiet master can end up genuinely weaker than everything around it.
Loudness Range Tells You Something the Average Cannot
Two files can share an integrated loudness and behave completely differently. Loudness range is the spread between the quiet passages and the loud ones.
Narrow is normal and correct for speech. A podcast that stays within a few units is easy to listen to in a car or on earbuds, because nothing disappears under road noise.
Wide is where trouble starts for spoken content. A wide range means the loud parts are comfortable and the quiet parts are inaudible in any real listening environment, which is what makes people reach for the volume control twice a minute. For music a wide range can be exactly the point, so the number is information rather than a verdict.
Sample Peak, and Why It Is Not True Peak
The peak reading here is sample peak: the largest value actually present in the decoded file.
True peak, which the standard also defines, reconstructs the waveform between samples and can read slightly higher, because a signal can overshoot between two sample points without either sample showing it. Measuring it properly needs oversampling.
So treat a sample peak close to zero as close enough to be a risk rather than a clean pass, and leave yourself headroom. This tool does not report true peak, and it says so rather than labeling one number as the other.
What It Cannot Tell You
Whether the audio sounds good.
Loudness is a measurement, not a judgment. A file can hit its target exactly and still be harsh, muddy, over compressed, or recorded in a bad room. None of that shows up in LUFS, because LUFS is answering one narrow question well rather than every question badly.
It also measures the file you give it. If you are checking a mix that will later be mastered, the number will change. Measure the thing you are actually going to publish.
Loudness Against the Other Audio Checks
Loudness and technical cleanliness are separate questions, and it is worth running both. The audio quality checker looks at clipping, average level in dBFS, noise floor and signal to noise, which is the “is this recording damaged” question. This page answers “is this the right volume for where it is going”.
A file can pass one and fail the other in either direction. Clean audio at the wrong loudness will be turned down by the platform. Correctly loud audio can still be clipped and hissy.
To fix what you find, the voice recording cleanup reduces background noise, and the audio cutter trims the dead air at the ends that gating ignores but listeners do not.
