Drop the audio or the finished video of an ad and read its integrated loudness, loudness range and true peak, then compare them with EBU R 128, ATSC A/85, Spotify and Apple targets. The file is decoded and measured on this device and never uploaded.
Drop an audio or video file
MP4 · MOV · M4A · MP3 · WAV · WebM. Decoded by your browser at 48 kHz.
The soundtrack is decoded, brought to the target and encoded again, while the picture is copied into the new file untouched. The new file is then decoded and measured again, and that measurement, not the plan, says whether it meets the target.
K weighting. A high shelf that models the head and a high pass filter, with the coefficients from BS.1770-5 Tables 1 and 2, so bass counts for less than it shows on a peak meter.
Channel weights. Left, right and centre count 1.0, the surround channels 1.41, and the LFE channel is left out, per BS.1770-5 Table 3.
Integrated loudness. The mean of 400 ms blocks overlapping by 75%, after an absolute gate at -70 LUFS and a relative gate 10 LU below the gated level, so silence and quiet passages do not drag the number down.
Momentary and short term. Ungated sliding windows of 400 ms and 3 s. The maximum short term reading is what EBU R 128 s1 limits for commercials.
Loudness range. The spread between the 10th and 95th percentile of short term readings above -70 LUFS and 20 LU below their mean, per EBU Tech 3342. It needs a minute or more of audio to mean much, which is why R 128 s1 does not use it for short form content.
True peak. The signal is oversampled four times with the 48 tap filter from BS.1770-5 Annex 2, which catches peaks that fall between samples. A peak meter reading of -1 dBFS can hide a true peak above 0 dBTP that clips in an AAC or MP3 encoder.
Delivery targets and where they come from
Pick the target in the box next to the drop zone. Every number below was read on the publisher's own page on 7 October 2026.
YouTube, TikTok, Meta and Instagram publish no target
None of the four publishes a loudness target for uploads or for ads in its help centre or ad specs. Figures quoted around the web come from measuring what the players do, not from the platforms. The meter therefore labels its social preset as practitioner guidance: -14 LUFS integrated with true peak at or below -1 dBTP is a common delivery level that survives normalization and lossy encoding, and nothing more official than that. If a media plan or a broadcaster names its own spec, that spec wins.
How the verdicts and the gain work
A reading passes when it is within the tolerance the target publishes. Within 1 LU beyond that it is a warning, which is our own band to separate "nearly there" from "remix it", not part of any standard. Further out, or with a true peak over the ceiling, it fails. Readings are compared at the 0.1 precision EBU Mode displays.
The gain is the change that lands the integrated loudness exactly on the target, and the meter shows where the true peak would end up. If it would cross the ceiling, gain alone cannot meet the target and a true peak limiter is needed. "Download normalized WAV" applies that gain and nothing else: no limiter, 24 bit PCM at 48 kHz, with a warning when samples would clip at full scale.
Loudness fixed inside the video
With AdCheckLab Pro, a video file gets a second way out: the fix goes into the file itself, so there is no WAV to lay back in an editor. The page works out the gain from the samples it measured. Where that gain would push the true peak past the target's ceiling, a true peak limiter joins it, set 1 dB under the ceiling because an AAC encoder can add a fraction of a dB of peak, and the gain is raised until the limited audio reaches the target again. The soundtrack is then decoded, processed and encoded as AAC (Opus for a WebM) with the WebCodecs encoder of your browser, and Mediabunny writes it into a new file next to the original video packets, copied byte for byte, so the picture loses nothing.
The new file is decoded and measured again with the same meter before you download it, and the before and after readings sit side by side. If the measurement says the target was missed, the page says so; it never reports a pass the new file did not earn. Encoding AAC needs a browser with a WebCodecs AAC encoder, such as current Chrome on a Mac, and the page asks your browser before it starts. Where the browser has none, or the picture uses a codec that cannot be copied into a new file, the page explains why and the normalized WAV stays available. Files up to 10 minutes can be fixed this way, and nothing is uploaded.
Tested against the EBU signals
The meter is checked on every build against the EBU minimum requirement signals, synthesised in the test suite: Tech 3341 cases 1 to 6 (stereo and 5.0 tones read -23.0 and -33.0 LUFS within ±0.1, including the gated sequences), 9, 10, 12 and 13 (short term and momentary maxima), 15 to 19 (true peak within +0.2 and -0.4 dB, including a quarter sample rate tone whose sample peak reads 3 dB low), and Tech 3342 cases 1 to 4 (loudness range 10, 5, 20 and 15 LU within ±1). Browsers resample every file to 48 kHz while decoding, which can move a true peak by a fraction of a dB on 44.1 kHz material.
Before and after the loudness check
For the picture, the Ad Spec Checker reads the video file itself against each placement's specs, including its audio codec and sample rate, and the Safe Zone Checker shows whether captions and logos sit under the platform interface. For the words around the ad, the Ad Copy Checker counts headlines and captions the way each platform does.
Questions about the Loudness Meter
No. The browser decodes the file with its own audio decoder and the measurement runs in a background worker on your device. The page keeps working offline once it has loaded, and the file name is never sent to analytics.
None of them publishes a number. Practitioners usually deliver around -14 LUFS integrated with true peak at or below -1 dBTP, which is what the social preset checks. For a TV buy in Europe use EBU R 128 s1, -23 LUFS with short term at or below -18 LUFS; in the US use ATSC A/85, -24 LKFS.
The waveform between two samples can rise above both of them. A sample peak meter only sees the samples, so it under reads, by as much as 3 dB for a tone at a quarter of the sample rate. The true peak is what a decoder or a digital to analogue converter actually reproduces, which is why every target limits it.
Gain moves loudness and peaks together. If the mix is quiet but peaky, raising it to the target pushes the peaks over the ceiling, and the meter says so before you download. Fix it with a true peak limiter in your editor, then measure the limited file again.
Yes, when the browser can decode the soundtrack, which covers MP4 and MOV with AAC audio and WebM with Opus in current Chrome, Edge and Firefox, and most of them in Safari. If a file will not decode, export the soundtrack as WAV and drop that.
Yes, with Pro. Only the soundtrack is processed and encoded again; the video packets are copied into the new file as they are, and the page checks that the first frame is identical byte for byte and that no frame is missing. The new file is then measured again, and the verdict you see is that measurement.
