Free tools for your next audio edit
Audio normalizer. Set the peak. Keep the balance.
Raise or lower a short recording to a chosen sample peak. See the measured level and required gain, compare the original with your normalized WAV, and download it. One gain changes the complete mix; stereo channels stay linked.
Peak normalizer
Free · localNo file selected. Audio stays in this browser.
Free with no account or audio upload. Peak normalization changes the whole mix’s level. It does not remove noise, repair clipping, level individual words, or match a LUFS target. Keep the original and download your result before leaving.
A deliberate level adjustment
How to normalize an MP3 or WAV
- 01 / MEASURE
Open a short recording
Choose mono or stereo MP3/WAV up to 50 MiB and 5 minutes. The browser decodes the file at 48 kHz. The report measures its highest sample peak across every channel, without uploading your audio.
- 02 / TARGET
Set the sample peak
Use a target from −24 to −0.1 dBFS. The default is −1 dBFS. Read the gain before applying it: a positive value boosts the complete recording, while a negative value lowers it.
- 03 / LISTEN
Compare and download
Prepare the WAV, then alternate between the original and normalized previews. Start at a low playback volume. Download the result before switching files or leaving; the tool does not save a session.
One gain across the recording
What peak normalization changes
Peak normalization adjusts a recording so its highest stored sample reaches a selected level. Every sample receives the same multiplier. Quiet passages retain their relationship to loud passages, and both stereo channels keep their relative balance. The tool measures the full decoded recording rather than estimating its peak from a waveform thumbnail.
gain = 10^(target dBFS ÷ 20) ÷ measured sample peakA decoded peak of 0.25 is about −12.04 dBFS. A −1 dBFS target requires about +11.04 dB of gain. A quieter right channel receives that same gain rather than being independently raised to the left channel’s peak.
Audacity’s Normalize reference also describes linking stereo channels to preserve their balance. This tool always links them and applies only gain; it does not include Audacity’s optional DC-offset removal.
Choose by the problem you hear
When normalizing audio helps
| Your recording | Useful starting step | Listen for |
|---|---|---|
| A quiet clip with modest peaks | Peak normalization can raise the whole clip | Hiss and room sound becoming more audible too |
| A clip that needs lower peaks for an edit | Choose a lower sample-peak target | The level in context with surrounding audio |
| Quiet dialogue with one loud impact | Inspect and edit the impact before deciding on gain | The impact setting the gain for the whole file |
| Alternating quiet and loud speakers | Use track gain, automation, or appropriate dynamics processing | Uneven sentences that constant gain leaves uneven |
| Speech covered by another sound | Evaluate noise reduction or source isolation | Missing words or artifacts in processed speech |
| A playlist requiring consistent perceived loudness | Use a loudness meter and an intentional LUFS workflow | Different loudness despite equal sample peaks |
Sample peak and perceived loudness
Use the right measurement for delivery.
This report uses dBFS sample peak. It does not measure RMS, LUFS, or true peak. A sparse recording and a dense mix can reach the same peak while sounding very different in loudness. Read the peak and loudness normalization guide before treating a peak target as a delivery requirement.
A target below 0 dBFS leaves space below the sample limit, but it is not a true-peak guarantee. Reconstructed peaks between samples, later effects, mixing, or lossy encoding can change the final peak. For a specified delivery, measure the final export with suitable loudness and true-peak tools.
Raising a noisy mix raises the wanted sound and the noise together. It does not improve their relative level. For a competing source, the audio cleanup chooser helps you select a separate workflow. Keep the original and assess any cleanup before a final level adjustment.
Export settings and practical limits
Know what the downloaded WAV contains.
The output is 48 kHz, 16-bit PCM WAV with the full decoded duration and original mono or stereo channel count. No trimming, edge fade, channel remix, automatic compressor, limiter, or metadata copy is applied. Browser decoding can resample the input; integer export quantizes the adjusted samples, so the final peak can differ slightly from the target.
The original file is unchanged. An MP3 becomes a WAV rather than a new MP3. That WAV can be larger than the compressed input. Both input and output must fit within 50 MiB; a long stereo export can exceed that limit even when its MP3 fits. Use the size calculator to plan an excerpt and the free trimmer to prepare it.
Support depends on the browser’s decoder. Try a standard PCM WAV if a recording cannot open. A silent decoded file has no peak to normalize. Samples already damaged by clipping remain damaged after lowering their level; gain does not reconstruct the missing waveform.
If only one side contains the useful audio, the channel splitter can extract it first. To check the resulting export’s stored settings and size, use the WAV inspector.
A few useful answers
Free audio normalizer questions
How do I normalize audio online for free?
Choose a supported mono or stereo MP3/WAV up to 50 MiB and 5 minutes. Read its decoded sample peak, choose a target from −24 to −0.1 dBFS, and prepare the normalized WAV. Preview the original and result, then download the 48 kHz, 16-bit PCM WAV. The output must also fit within 50 MiB.
Does this normalize loudness to a LUFS target?
No. It uses the largest absolute decoded sample across all channels to set one constant gain. It measures sample peak in dBFS, without LUFS, RMS, or true-peak measurement. Equal sample peaks do not ensure equal perceived loudness between recordings.
Will normalization remove background noise or repair clipping?
No. Gain raises or lowers every sound together, including hiss, room noise, and existing distortion. It does not recover a clipped waveform, reduce noise relative to speech, or make quiet and loud words more even. Choose a different treatment when those are the problems.
Are left and right channels normalized separately?
No. Both channels receive the same gain, calculated from the highest sample peak across either channel. This preserves their relative balance. A quieter channel can therefore finish below the target. A mono recording remains mono.
What happens to my MP3 or WAV?
The original file stays unchanged. The browser decodes it at 48 kHz, then the tool applies gain and exports 16-bit PCM WAV with the full decoded duration and channel count. It does not encode MP3, retain original metadata, remove DC offset, or preserve the source bit depth. Codec support varies by browser.
Is my recording uploaded, and do I need credits?
No audio is uploaded, and normalization, preview, and WAV download require no account or credits. The tool reads audio in this browser and clears the session when you leave or reload. Saturalabs’ separate AI source-isolation service has its own account and download-credit flow.
Why does a silent file have no normalization result?
A decoded recording with only zero-valued samples has no nonzero peak. There is no finite gain that brings silence to the target, so the tool explains the result and leaves export disabled. For a very quiet recording, a large boost will also raise any noise already present.
After preparing your audio