This LUFS meter measures an audio or video file the way streaming services, broadcasters and a DAW's loudness meter do: integrated loudness in LUFS over the whole file with the gating of ITU-R BS.1770-4, the true peak in dBTP at four-times oversampling, the loudness range in LU per EBU Tech 3342, and the momentary and short-term maxima. Open an MP3, WAV, FLAC, AAC, M4A, OGG or Opus file, or an MP4, MOV or WebM video, and the browser decodes the audio track on your device; a chart shows where the loud and quiet passages fall, and a table gives the gain in dB that would bring the file to each published target, with the true peak it would then reach. Nothing is uploaded: the measurement is JavaScript running in your browser, checked against the EBU's own test signals.
How to use the meter
- Open the file. Choose a file or drop one on the box. For a video the browser decodes the audio track; for a song, a podcast episode or a voice-over it decodes the file itself. Files the browser cannot decode (some MKV and AVI containers, camera-raw audio, DRM-protected tracks) give an error, and exporting the audio as WAV, FLAC, MP3 or M4A fixes it.
- Wait for the bar. Decoding takes a moment, then the measurement runs at about a minute of stereo audio per half second on a laptop; a ten-minute file is done in a few seconds, an hour-long one in about half a minute. A one-hour file needs about 700 MB of memory once decoded, which a phone may not have.
- Read the figures. Integrated loudness is the number specifications quote; true peak is what the ceilings are set against; loudness range tells you how much the level moves; the short-term and momentary maxima show the loudest 3 seconds and loudest 400 ms. The chart draws the short-term loudness over time with the integrated value and the -14 LUFS line, and the table under it lists the gain to each target. Copy the figures as text puts a plain-text summary on the clipboard for a delivery note or a message to the mix engineer.
What the numbers mean
Integrated loudness (LUFS) is one number for the whole file. The signal is K-weighted (a high shelf that lifts the upper mids and treble the ear is sensitive to, and a high-pass that removes the sub-bass it is not), the mean square is taken over 400 ms blocks that overlap by 75%, and the blocks are gated twice: anything under -70 LUFS is dropped as silence, then anything more than 10 LU under the average of what is left is dropped as background. The average of the surviving blocks is the integrated loudness. The gating is why a podcast with long pauses and a podcast without them measure alike, and why a file cannot be made to measure quieter by adding silence at the end.
Short-term loudness is the same measurement over a sliding 3 second window, ungated, and momentary loudness over 400 ms. The maxima of each are what some delivery specifications limit (EBU R 128 s1 sets a maximum short-term loudness for short-form content such as advertisements). The chart on this page plots the short-term value once a second, so a shouted intro, a loud ad read or a quiet interview answer shows up as a shape, not a number.
Loudness range (LRA, in LU) describes how far the short-term loudness moves across the file, measured per EBU Tech 3342: the short-term values are gated at -70 LUFS and at 20 LU under their average, and the range is the distance between the 10th and the 95th percentile of what remains, so a single explosion or a few seconds of silence do not set it. A heavily compressed pop master measures 3 to 8 LU, a jazz trio or a classical recording 12 LU and more, a feature film mix 15 to 20 LU. A narrow range suits earphones and phone speakers; a wide one suits a quiet room and a proper system, and sounds like a volume problem on a train.
True peak (dBTP) is the highest level the reconstructed waveform reaches, not the highest sample. Between two samples the analogue signal a converter or a lossy encoder rebuilds can overshoot both of them, so a file whose samples stop at -0.3 dBFS can clip at +0.5 dB on playback. BS.1770-4 measures it by oversampling four times (for 48 kHz audio) and reading the largest interpolated value; this page does the same and shows the plain sample peak beside it so you can see the gap. Specifications set their ceilings in dBTP, usually -1 or -2 dBTP, precisely so that the encoder on the way to the listener has room.
Published loudness targets
The figures below are the ones each organisation publishes, with the page they come from and the date it was read; they change occasionally, so check the source before a delivery. The table in the result applies them to your file: the gain is target minus measured, and the "true peak after gain" column adds that gain to the measured true peak, which tells you whether a plain gain change reaches the target or a limiter has to come first.
| Target | Integrated loudness | True peak ceiling | Notes and source |
|---|
| Spotify | -14 LUFS (Loud setting -11, Quiet -19) | -1 dBTP; -2 dBTP if the master is louder than -14 LUFS | Normalisation per ITU-R BS.1770; quieter masters are raised, but only as far as 1 dB of true-peak headroom allows (a -20 LUFS master peaking at -5 dBTP is lifted to -16, not -14). Off in the web player and on some third-party devices. Spotify for Artists: loudness normalization, read 2026-10-10. |
| Apple Podcasts | -16 LKFS, ±1 dB | -1 dB FS true peak | Measured per ITU-R BS.1770; apply before encoding. Apple Podcasts audio requirements, read 2026-10-10. |
| Apple Music | not published | at least 1 dB of headroom | Apple's Digital Masters document says Sound Check measures loudness per BS.1770 and stores it in the file's metadata, warns that oversampling can clip a 0 dBFS master, and asks for 1 dB of headroom; it names no LUFS figure. Apple Digital Masters (PDF), read 2026-10-10. |
| EBU R128 (European broadcast) | -23.0 LUFS; ±1.0 LU for live programmes, ±0.2 LU measurement tolerance in QC | -1 dBTP (±0.3 dB measurement tolerance) | The recommendation that defined LUFS, LU and the gating; loudness range is reported, not limited. EBU R 128-2023, version 5, November 2023, read 2026-10-10. |
| ATSC A/85 (US television) | -24 LKFS, ±2 dB | -2 dBTP | The US broadcast practice the CALM Act references; LKFS is the same unit as LUFS. ATSC A/85, Techniques for Establishing and Maintaining Audio Loudness for Digital Television. |
| YouTube | not published | not published | The player turns loud uploads down and reports the offset as "content loudness" in the Stats for nerds panel; a quieter upload shows a negative figure there and plays at 100%, so it is not raised. Measurements by mastering engineers put the reference near -14 LUFS; that is a community figure, not a Google one. |
Two patterns follow from the table. Music and podcast platforms sit between -14 and -16 LUFS with a -1 dBTP ceiling, so a master at -14 LUFS and -1 dBTP is played at its own level on Spotify and turned down by about 2 dB by Apple Podcasts, while a master at -8 LUFS is turned down 6 to 8 dB everywhere and gains nothing from its loudness except the limiting it took to get there. Broadcast sits 8 to 10 LU lower with wider tolerances and more dynamic range, which is why a television mix sounds quiet when it is posted unchanged to YouTube, and a YouTube video sounds like an advert when it is cut into a programme.
LUFS and dB
A loudness unit is a decibel: 1 LU = 1 dB, and LUFS values are decibels relative to full scale, the same reference as dBFS. What differs is what the decibel is measuring. dBFS peak describes individual samples; RMS describes the average power over a window with no weighting; LUFS describes the K-weighted, gated average that BS.1770 defines. So "convert LUFS to dB" has a plain answer when the question is about gain (a file at -18.3 LUFS needs +4.3 dB to reach -14 LUFS, and the box above the guide does the subtraction and adds the true peak) and no answer when the question is about peak level: two files can both peak at -1 dBFS and sit 10 LU apart, because one is dense from start to finish and the other breathes. The chart on the result shows which kind of file you have.
Reaching a target
If the true peak after gain stays under the ceiling, a gain change is the whole job: raise or lower the file by the figure in the table, and its dynamics are untouched. If the true peak after gain would pass the ceiling, the peaks have to be limited first, or the loud moments clip once the gain is applied. The gap between the two numbers is the limiting that is needed: a file at -20 LUFS with a -4 dBTP true peak needs +6 dB to reach -14 and would then peak at +2 dBTP, so about 3 dB of true-peak limiting has to come first to land at -1 dBTP. Limiting raises the loudness of what remains, so measure again after it.
Every DAW has a loudness meter and a true-peak limiter; for a finished video, ffmpeg's loudnorm filter does both in one pass and leaves the picture untouched:
ffmpeg -i in.mp4 -af loudnorm=I=-14:TP=-1:LRA=11 -c:v copy out.mp4
I is the integrated target in LUFS, TP the true-peak ceiling in dBTP, LRA the loudness range it may compress towards; -c:v copy copies the video stream without re-encoding it, so only the audio changes. Single-pass loudnorm adapts as it goes and can land a little off target on short files; measuring the output here and running a second pass with the printed figures closes the gap. For a podcast or a music file, drop -c:v copy and write an MP3, M4A or WAV.
How the measurement works
The page implements ITU-R BS.1770-4 directly in JavaScript. Each channel goes through the two K-weighting filters (the high shelf centred near 1.7 kHz with about +4 dB above it, and the second-order high-pass near 38 Hz; the coefficients are computed for the file's sample rate and match the standard's 48 kHz table to six decimals), the mean square of each 100 ms is kept, and 400 ms blocks are formed from four of them with 75% overlap. Channels are summed with the standard's weights: 1 for left, right and centre, 1.41 for the surround pair, and the LFE channel is left out of a 5.1 file. The absolute gate drops blocks under -70 LUFS, the relative gate drops blocks more than 10 LU under the average of the rest, and the integrated loudness is -0.691 dB plus ten times the log of the surviving blocks' mean power, so that a full-scale 1 kHz tone in one channel reads -3.01 LUFS and a stereo pair of them 0 LUFS.
The true peak uses a four-times oversampling polyphase filter (12 taps per phase, 48 in all, windowed sinc) at sample rates under 96 kHz, two-times from 96 to 192 kHz, and the sample value itself above that, as the standard's Annex 2 lays out, which keeps the error on the EBU test tones under 0.1 dB. The loudness range follows EBU Tech 3342: 3 second windows every 100 ms, the -70 LUFS absolute gate, a -20 LU relative gate on the mean power, and the 10th-to-95th percentile span. Before the page was published, the code was checked in Node.js against the cases of EBU Tech 3341 and 3342 that can be synthesised: the -23 and -33 dBFS stereo tones read -23.0 and -33.0 LUFS, the gating sequences (-36/-23/-36 dBFS and the version with -72 dBFS leaders) read -23.0, the 5.1 weighting case reads -23.0, the loudness-range sequences read 10, 5, 20 and 15 LU, and the quarter-, sixth- and eighth-sample-rate true-peak tones read -6.0 dBTP within 0.1 dB, all inside the tolerances the EBU sets.
One honesty note on the decode step. The browser's Web Audio API decodes the file with the codecs the browser ships, and this page asks it to deliver the audio at 48 kHz, so a 44.1 kHz MP3 is resampled on the way in. Resampling does not change loudness, and the true peak is measured on the 48 kHz signal with the oversampling the standard prescribes for that rate, so the figures match a 44.1 kHz meter within the tolerances above; it does mean the sample rate shown in the result is the measurement rate, not the file's.
YouTube and loudness
YouTube's help pages name no loudness target. What the player does is visible in Stats for nerds (right-click the player): a line such as "Volume / Normalized 100% / 61% (content loudness 4.3dB)" means the upload measured 4.3 dB louder than the reference and is being played 4.3 dB down; "content loudness -2.1dB" with "100% / 100%" means the upload is quieter than the reference and is played as it is, so quiet uploads are not raised. Measurements by mastering engineers who have compared that readout with their own meters put the reference near -14 LUFS, and that is the figure most mix guides for YouTube quote; it is their measurement, not Google's specification, and the readout on your own upload is the authoritative check. The practical reading is the same either way: a video mixed near -14 LUFS with a true peak under -1 dBTP plays at full volume next to everything else, a louder one is turned down to the same place, and a quieter one stays quiet. Google's upload encoding recommendations cover the audio format (AAC-LC, 48 kHz, 128 kbps mono, 384 kbps stereo, 512 kbps 5.1) but not the level; the video bitrate calculator has those figures with the video bitrates beside them.
What the meter does not do
It does not change the file: there is no normalise or export button, because applying gain or limiting properly belongs in the editor that made the file, and the ffmpeg line above does it for a finished video. It does not read the loudness of a YouTube video from its URL; the audio stream of a video on youtube.com is not readable cross-origin from a page, and the Stats for nerds readout is the published way to see how the player treats an upload. It reports integrated, short-term and momentary loudness, LRA and true peak, not dialogue-gated loudness (the dialogue-intelligence measure some streaming film specifications use) or per-channel figures. And it measures what the browser decodes: a DRM-protected track, a container the browser does not open, or a file too large for its memory give an error rather than a number.
Frequently asked questions
What is LUFS?
LUFS stands for Loudness Units relative to Full Scale. It is the loudness measure defined in ITU-R BS.1770: the signal is filtered to approximate how loud the ear perceives it (K-weighting), averaged over 400 ms blocks, and the quiet blocks are gated out so that silence and pauses do not drag the figure down. The result is a single number for a whole file, the integrated loudness. 0 LUFS is the loudness of a full-scale 1 kHz tone; music masters sit around -8 to -14 LUFS, podcasts around -16, television around -23 or -24. LKFS is the same unit under its American name, and 1 LU is one decibel.
What is the difference between LUFS and dB?
The scale is the same: 1 LU equals 1 dB, so raising a file by 3 dB raises its loudness by 3 LU. The difference is what is measured. dBFS describes the level of samples (a peak at -1 dBFS is one decibel under full scale); LUFS describes perceived loudness over time after K-weighting and gating. Two files can both peak at -1 dBFS and differ by 10 LUFS, because one is dense and the other has quiet passages. The gain to reach a target is simply target minus measured: a -18.3 LUFS file needs +4.3 dB to reach -14 LUFS.
What LUFS should a YouTube video be?
YouTube publishes no loudness target. Its player turns loud content down and shows the offset in the Stats for nerds panel as a content loudness figure in dB; a video quieter than the reference shows a negative figure there and plays at 100% volume, so quiet content is not raised. Independent measurements by mastering engineers place the reference around -14 LUFS, and that figure is widely used, but it is a community measurement, not a Google specification. Mixing to about -14 LUFS integrated with a true peak under -1 dBTP means the player neither turns the video down nor leaves it quieter than its neighbours.
What is true peak and why is it higher than the sample peak?
The sample peak is the largest sample value in the file. The true peak estimates the largest value of the analogue waveform the samples describe, which can lie between two samples. A digital-to-analogue converter or a lossy encoder reconstructs that waveform, so a file whose samples never exceed -0.1 dBFS can still clip on playback. ITU-R BS.1770-4 measures it by oversampling four times and reading the largest interpolated value, in dBTP. Streaming and broadcast specifications set their ceilings in dBTP, usually -1 or -2.
What is a good loudness range (LRA)?
Loudness range measures how much the short-term loudness varies across a file, in LU, after dropping the quietest and loudest extremes. A dense pop master typically measures 3 to 8 LU, a dynamic acoustic recording 10 to 15 LU, a film mix 15 to 20 LU or more. There is no right value, only a fit: a wide range suits a cinema and a quiet room, a narrow one suits earphones on a train and a phone speaker. EBU R128 names no maximum but suggests checking it; a podcast with an LRA above about 10 LU usually has sections listeners will reach for the volume control on.
Why does my DAW meter show a different number?
Three settings explain most gaps. Integrated versus short-term: this page reports the integrated loudness of the whole file; a meter that shows the current short-term or momentary value reads higher in loud passages and lower in quiet ones. Gating: BS.1770-4 gates at -70 LUFS absolute and 10 LU under the ungated average, so a meter set to the older -8 LU relative gate or to no gate reads a little lower on material with pauses. True peak oversampling: a meter at 2x or 8x oversampling lands a few tenths of a dB from a 4x figure on bright, dense material. Measured on the EBU test signals, this page lands within 0.1 LU of the specified values.
How do I make a file louder or quieter to hit a target?
Apply the gain the table shows: target minus measured. If the true peak after that gain stays under the ceiling, plain gain is enough and the file keeps its dynamics. If the true peak would pass the ceiling, a limiter has to catch the peaks first, or the file reaches the target with clipped transients. In ffmpeg the loudnorm filter does both for a target of -14 LUFS and a -1 dBTP ceiling: ffmpeg -i in.mp4 -af loudnorm=I=-14:TP=-1:LRA=11 -c:v copy out.mp4. Measure the output again, since a limiter changes the loudness it is set against.
Is the file uploaded?
No. The page opens the file with the browser's own audio decoder, measures it in JavaScript on your device and discards it when you close the tab. The only network traffic is the page itself. A one-hour file needs about 700 MB of memory in the browser once decoded, so very long files can fail on a phone or a small laptop; exporting the audio track alone, or a shorter section, gets round that.
