LUFS loudness meter and normalizer (EBU R128 / BS.1770) for streaming and podcasts
Integrated loudness, LRA and true peak per BS.1770, then a normalized WAV, no upload.
- Instant
- Free
- Private (processed locally)
- No sign-up
Measure the LUFS loudness and true peak of an audio file
This online LUFS meter implements ITU-R BS.1770-4 with EBU R128 gating, the same method as mastering plug-ins and broadcast meters: integrated, momentary and short-term loudness, loudness range (LRA) per EBU Tech 3342, true peak by 4× oversampling and sample peak, per channel. It then tells you what Spotify, Apple Music, YouTube, Apple Podcasts or television will do with your file: by how many decibels it will be turned down or up, and whether its peaks exceed the ceiling.
You can also normalize the file to a target (−14 LUFS for Spotify, −16 for a podcast, −23 for TV or any value) with a true-peak ceiling, either by reducing the gain or with a true-peak limiter, and download a 16, 24 or 32-bit WAV whose values have been measured again. Everything happens in your browser, in a Web Worker: no file is uploaded, even for an hour of audio.
-
Choose an audio file
Click “Choose an audio file” or drop an MP3, WAV, FLAC, OGG, M4A or video file on the page. The tool decodes it, at its original sample rate when possible, and shows duration, channels and format; a synthetic demo is measured as soon as the page opens.
-
Pick the target and the ceiling
Select a platform in “Target” (Spotify, Apple Music, YouTube, Apple Podcasts, EBU R 128…) or type a custom LUFS value, then set the “True-peak ceiling” (−1 dBTP by default). The big readout, the verdict and the table update immediately.
-
Read the results
Integrated loudness, loudness range, momentary and short-term maxima, true peak per channel and the gating threshold are listed; the graph shows the loudness over time and the waveform shades the parts excluded by the gates. The platform table says by how many dB each service will turn the file up or down.
-
Normalize and download the WAV
Choose the WAV format (16-bit with dither, 24-bit or 32-bit float), decide what happens when the gain would push the true peak above the ceiling (reduce the gain or apply the true-peak limiter), then click “Normalize and download the WAV”. The exported samples are measured again and the real values are displayed.
How the BS.1770 / EBU R128 meter works
The meter implements ITU-R BS.1770-4 (the algorithm is unchanged in BS.1770-5) with EBU R128 gating, exactly as broadcast meters and libraries such as libebur128 do, inside a Web Worker so that even an hour of audio does not freeze the page:
- K-weighting: a high shelf that lifts the treble by 4 dB (transition around 1.7 kHz) and a 38 Hz high-pass, with coefficients computed for the file’s own sample rate; at 48 kHz they reproduce the table of the standard to 14 decimals.
- Gating blocks: mean square per channel over 400 ms blocks every 100 ms (75% overlap), channel weights 1.0 for L, R, C and 1.41 for Ls, Rs, LFE ignored; block loudness = −0.691 + 10 log10 of the weighted sum.
- Integrated loudness: blocks under −70 LUFS are dropped, then blocks more than 10 LU below the mean of the remaining ones; the result is the mean of what is left.
- Momentary and short-term: the 400 ms and 3 s windows with their maxima; loudness range per EBU Tech 3342 from the short-term values (every 100 ms), gated at −70 LUFS and −20 LU, between the 10th and 95th percentiles.
- True peak: 4× oversampling with the 48-tap, 4-phase FIR filter printed in Annex 2 of BS.1770-4 (2× from 96 kHz, none from 192 kHz), per channel, in dBTP.
- Normalization: gain = target − integrated loudness. If that pushes the true peak above the ceiling, the tool either lowers the gain or runs a look-ahead brickwall limiter (5 ms look-ahead, 100 ms release, detection on the oversampled signal), then measures the exported samples again.
Loudness targets of the main platforms (2026)
| Platform | Target | True peak | Notes |
|---|---|---|---|
| Spotify | −14 LUFS | −1 dBTP (−2 if louder than −14) | Official: Loud −11, Normal −14, Quiet −19 settings; quiet tracks are raised only until the true peak reaches −1 dB |
| Apple Music and Apple Podcasts | −16 LUFS (±1 for podcasts) | −1 dBTP | Sound Check plays at −16 dB; Apple Podcasts measures per BS.1770-5 |
| YouTube | ≈ −14 LUFS | −1 dBTP recommended | Not published; measured by engineers, loud videos are only turned down |
| Amazon Music, Tidal, Deezer | ≈ −14, −14 and −15 LUFS | −1 to −2 dBTP | Reported values, not published by the platforms; Tidal normalizes per album |
| EBU R 128 (TV, Europe) | −23 LUFS (±0.5 LU; ±1 LU live) | −1 dBTP | Official recommendation, revised in 2020 and 2023; supplement R 128 s1 for short-form content |
| ATSC A/85 (TV, USA) and Netflix | −24 LKFS ±2 dB; Netflix −27 LKFS ±2 LU on dialogue | below −2 dBTP | A/85:2026-07; Netflix measures dialogue with BS.1770-1 gating, not comparable |
Checked in October 2026 in the documents published by each platform where they exist. The actual adjustment also depends on the listener’s settings.
What a LUFS meter is used for
- Checking a master before distribution: knowing whether Spotify or Apple Music will turn it down and by how much, and whether the peaks leave the headroom needed for AAC or Ogg encoding.
- Leveling podcast episodes to the same loudness (−16 LUFS, −1 dBTP) so listeners never touch the volume between episodes.
- Preparing the sound of a video for YouTube, a European TV channel (−23 LUFS) or an American one (−24 LKFS), with the true-peak control broadcasters demand.
- Comparing two versions of a mix by loading them one after the other: at equal loudness only the music changes, without the “louder sounds better” bias.
- Checking a file you received from a mastering service, a voice-over artist or an automatic tool before publishing it.
Common mistakes and how to fix them
- Master too loud for Spotify (−8 LUFS, say): it will be turned down by 6 dB and sound more squashed than its neighbors; aim for −14 to −10 LUFS while keeping dynamics.
- Peaks above −1 dBTP although the samples stay under 0 dBFS: AAC or MP3 encoding will clip; export with the true-peak limiter or lower the gain.
- Mono podcast reading 3 LU too low: that is how BS.1770 works, not a bug; aim the target at the file as it will be delivered, mono or stereo.
- Normalized file quieter than the target: the true-peak ceiling limited the gain (“Reduce the gain” mode); pick the limiter or accept the displayed level.
- The browser cannot decode the file (some browsers lack AAC codecs, WMA, rare formats): convert it to WAV or FLAC and try again.
Privacy: the audio is decoded, measured and normalized in your browser; nothing is uploaded or stored. Limits: accepted formats depend on your browser’s decoders; files longer than 90 minutes are refused and a warning appears above 45 minutes (decoded audio takes about 23 MB per minute in 48 kHz stereo). Measured values follow BS.1770-4 with 4× oversampling: other meters with longer filters can read a slightly higher true peak on very bright signals. Dialogue-gated (Netflix) and RMS-based (ACX) specifications are not equivalent to integrated loudness. The microphone mode only shows momentary and short-term loudness.
Frequently asked questions
What is integrated loudness in LUFS?
LUFS (loudness units relative to full scale) is the loudness scale of ITU-R BS.1770 and EBU R128. The integrated value sums up the whole file after K-weighting, a filter that mimics how the ear weights frequencies, and after gating that ignores silence and very quiet passages. One LU equals one decibel: a track at −10 LUFS is 4 dB louder than one at −14 LUFS, and this is the number streaming services use to turn each track up or down.
How loud should a track be for Spotify and YouTube?
Spotify normalizes to −14 LUFS and recommends a true peak below −1 dBTP, or below −2 dBTP if the master is louder than −14 LUFS; listeners can also pick Loud (−11) or Quiet (−19). YouTube publishes nothing, but engineers measure a reference of about −14 LUFS that only turns loud videos down. A master around −14 LUFS with peaks under −1 dBTP therefore plays at its own level on both services, while a louder master is simply attenuated.
What loudness should a podcast have?
Apple Podcasts asks for about −16 LKFS (the same unit as LUFS) with a tolerance of ±1 dB and a true peak no higher than −1 dBFS, measured per ITU-R BS.1770-5. Spotify publishes −16 LUFS ±1.5 and −2 dBTP for podcast ads, but its help pages give no target for episodes themselves. A spoken-word episode at −16 LUFS with peaks under −1 dBTP is a safe choice everywhere; remember that a mono file reads 3 LU lower than the same signal duplicated in stereo.
Why does a mono file read 3 LU lower than in stereo?
BS.1770 adds the energy of every channel with a weight of 1.0 for left, right and center. A 1 kHz sine at −23 dBFS in a single mono channel measures −26 LUFS, and the same sine in both channels of a stereo file measures −23 LUFS, because two channels carry twice the energy. The meter applies the standard literally and does not add 3 dB to mono files; when you export with the mono option, the tool measures the downmix again before normalizing.
What is the difference between true peak and sample peak?
The sample peak is the largest sample value in the file; the true peak estimates the highest point of the analog waveform between samples, by oversampling four times with the FIR filter of BS.1770-4 Annex 2. Peaks between samples can exceed the samples by 1 dB or more, and lossy encoders (AAC, MP3, Ogg Vorbis) add their own overshoot, which is why Spotify and Apple ask for 1 dB of headroom at −1 dBTP. A file whose samples never exceed −0.3 dBFS can still clip after encoding.
Why does this meter show different values than my plug-in or ffmpeg?
Integrated loudness agrees with ffmpeg’s ebur128 filter within 0.1 LU on all our test files. The loudness range depends on the hop between 3 s windows and on the percentile method, so step-like test signals can differ by a few tenths of an LU. True peak depends on the interpolation filter: this tool uses the 4× FIR of the standard, while ffmpeg resamples to 192 kHz with its own filter, which can read 0.5 to 1 dB higher on white noise or on a tone at exactly a quarter of the sample rate.