How to use the audio normaliser
- Add one audio file, or a video whose sound you want to level.
- Under Normalise by, pick Peak level or Loudness (LUFS).
- For peak, set Target peak (−20 to 0 dBFS, default −1). For loudness, choose a Target loudness and leave Keep peaks at or below −1 dBFS ticked unless you have a reason not to.
- Pick Save as (WAV by default) and press Normalise.
What it does and when to use it
Recordings made on different days or devices rarely come out at the same level. Normalising brings each file to a known reference, so listeners do not reach for the volume between episodes. Peak mode suits sound effects and samples that must not clip; loudness mode suits speech and music, because it follows how loud the audio sounds rather than one stray spike.
To join several clips and level them in one pass, the audio merger has a “Match loudness of each clip (−16 LUFS)” option. If you already know the exact change you want, the audio volume changer applies a fixed dB gain.
How it works
Decoding and processing happen in a Web Worker on your device.
- Peak: gain (dB) = target − current sample peak. A file peaking at −9 dBFS with a −1 dBFS target gets +8 dB.
- Loudness: the tool measures integrated loudness per ITU-R BS.1770-4: K-weighting filters, 400 ms blocks, an absolute gate at −70 LUFS and a relative gate 10 LU below the first average. Gain = target − measured. If the ceiling box is ticked and the peak would pass −1 dBFS, the gain is reduced to stop there.
Worked examples
Results from the tool on test tones:
- Peak, −1 dBFS target, file peaking at −9 dBFS: Gain applied 8 dB, Peak after −1 dBFS.
- Loudness, −16 LUFS, a quiet 1 kHz tone: Measured −29 LUFS, Gain applied 13 dB, Loudness after −16 LUFS, Peak after −13 dBFS.
- Loudness, −14 LUFS on a quiet track with sharp clicks: Measured −25.9 LUFS, but the clicks reach −1 dBFS after only 1.9 dB, so Loudness after is −24 LUFS and a warning explains the cap. With the ceiling unticked the full 11.9 dB is applied and the clicks are cut off at full scale.
Limits and tips
- Unticking the ceiling can clip. When the gain pushes peaks past full scale, Peak after shows the level with “(clips)”, Loudness after is marked “before clipping”, and a warning explains that the format clips those peaks (32-bit float WAV keeps them).
- Sample peak, not true peak. Allow a little headroom if you will convert to MP3 or AAC afterwards.
- Long files are decoded into memory in full; on a phone, shorten them first with the audio trimmer.
- MP3 output, and input formats such as AIFF, AMR or WMA, need the optional ffmpeg fallback, a one-time download of about 30 MB from cdn.jsdelivr.net. Your audio still stays on the device.
Frequently asked questions
Peak or loudness, which should I pick?
Peak normalising only guarantees the highest sample lands on your target; two files can still sound very different in volume. Loudness normalising measures how loud the whole file feels and is the better choice for podcasts, voice-overs and anything that plays next to other audio.
Which LUFS target should I use?
The menu offers −14 LUFS (common for music streaming), −16 LUFS (common for stereo podcasts), −19 LUFS (mono podcasts) and −23 LUFS (EBU R128 broadcast). If a platform publishes its own target, use that.
Why did my file not reach the LUFS target?
With 'Keep peaks at or below −1 dBFS' ticked, the gain stops when the loudest sample reaches −1 dBFS. Files with sharp drum hits or clicks hit that ceiling early. A warning shows the gain that was used; a limiter in a desktop editor is needed for the full target.
Does normalising change the dynamics?
No. One fixed gain is applied to every sample, so the gap between quiet and loud parts stays the same. It is not a compressor.
Why does it say the audio is too short or too quiet?
Loudness is measured in 400 ms blocks and ignores anything below −70 LUFS, so clips shorter than 0.4 s or almost silent files cannot be measured. Use Peak level for those.
How do we know nothing is uploaded? Test it yourself on the privacy proof page.