Audio Normalizer
Normalize the volume of an audio file to a target peak or loudness level and download a 16-bit WAV.
Drag & drop an audio file here, or click to browse
No uploads. Your audio stays on your device.
The solid wave is the normalized result; the faint outline behind it is the original level.
How it works
Drop an MP3, WAV, M4A, OGG, FLAC or WebM file and the Web Audio API decodes it right here in the page — nothing is uploaded, so podcast takes, voice memos and unreleased mixes never leave your device. Every sample is scanned to find the true peak (the single loudest sample) and the RMS level (the average energy, which is much closer to how loud a track actually sounds), both shown in dBFS where 0 dBFS is the digital maximum.
Peak level normalizing lifts the whole file so its loudest moment lands on your target — the safe default of −1 dBFS keeps a little headroom for encoders. Average loudness (RMS) matches perceived volume instead, which is what you want when several clips need to sit at the same level; −16 dBFS is a sensible starting point for spoken word. One single gain factor is applied to every sample, so nothing is compressed or squashed and the dynamics stay intact. Leave Prevent clipping ticked and the gain is capped so no peak ever crosses 0 dBFS; untick it and any samples that would overshoot are reported and hard-limited. The result plays back instantly and downloads as an uncompressed 16-bit WAV at the original sample rate and channel count.
Frequently asked questions
What does normalizing audio actually do?
Normalizing measures the level of the whole file and then applies one single gain factor to every sample so the file lands on the level you asked for. Peak mode looks at the loudest sample and lifts it to your target (−1 dBFS is a safe default that leaves a little headroom for encoders); average loudness (RMS) mode matches perceived volume instead, which is what you want when several clips must sit at the same level. Because it is one constant gain, nothing is compressed or squashed — the quiet parts stay quiet relative to the loud parts and the dynamics are untouched.
How do I make a quiet recording louder without distorting it?
Drop the file in, leave the mode on Peak level with a target of −1 dBFS and keep "Prevent clipping" ticked. The tool then applies the largest gain that still keeps every peak below 0 dBFS, and it tells you exactly how much gain that was. If you push for a loudness target that would need more gain than the peaks allow, the gain is capped and a note explains why; unticking the clipping guard is possible, but the tool will report how many samples were clipped so you know the file is distorting.
Is my audio uploaded anywhere?
No. The file is decoded with the Web Audio API inside this tab, the levels are measured from the decoded samples and the new WAV is built in memory — no server, no account and no link that could be shared by accident. That matters for interviews, voice notes and unreleased music. The download is an uncompressed 16-bit WAV at the original sample rate and channel count, so there is no second round of lossy compression.