Change Volume

Raise or lower the gain of an audio file by a set amount, without sending it anywhere.

Runs on your deviceAudioWorkingNo account needed

What this tool does

Change Volume decodes your file to raw samples, multiplies every one of them by a single constant factor, and writes a new file from the result. That factor is whatever you asked for: 150% is a gain of about +3.5 dB, 200% is exactly twice the amplitude and a shade over +6 dB, and a figure typed in decibels is converted the same way. Because one number is applied across the whole file, the balance between its quiet passages and its loud ones is left as it was — this is a fader, not a compressor.

Prevent clipping is on by default, and what it does is worth understanding before you rely on it. The loudest sample in the file is measured first; if the gain you asked for would push that sample past full scale, the gain is reduced until the peak lands just underneath instead, and the result line says it was pulled back and from what figure. The consequence is that a track already mastered close to 0 dBFS cannot be made meaningfully louder here — the pull-back cancels nearly the whole boost. Getting an already-loud track louder is the job of a limiter, which squashes the peaks so everything else can come up, and nothing in this tool does that. Switching the protection off applies your figure as typed, and anything past full scale is then flattened on the way into a 16-bit or 24-bit WAV or an MP3, which is audible distortion; the count of samples that crossed the ceiling is shown as a badge beside the result. A 32-bit float WAV is the exception, since it stores over-scale samples as they are and leaves the clipping to whatever plays it back.

Nothing about this job leaves the tab it starts in. Your file is decoded with the Web Audio API, edited as floating-point samples and written out again here, so no audio is transmitted. A WAV result needs nothing beyond the page itself and finishes with the network cut; an MP3 result fetches the LAME encoder from this site the first time you ask for one, so that first MP3 does want a connection, even though what crosses it is a script coming down rather than your audio going up. Two further limits follow from working in a browser rather than on a server. Output is WAV or MP3 only, because no browser hands a page an AAC, FLAC or OGG encoder — "Match the input where possible" keeps an MP3 as an MP3 and turns everything else into a WAV. And the file is rebuilt from samples upward, so ID3 tags and embedded cover art do not survive the trip.

How to use it

  1. Load the track you want to re-level. MP3, WAV, M4A/AAC, FLAC and OGG all work as far as your browser can decode them, and WAV is read by a built-in RIFF reader instead. Add several and each is processed on its own with the same settings, though the summary line and the badges under it end up describing whichever file was finished last.
  2. Set the level by percentage or in decibels. The percentage slider runs from 0 to 400% in steps of five and starts at 150%, where 100% leaves the file alone and 0% silences it. The decibel field starts at +6, and its arrows step half a decibel at a time between −60 and +30; a figure typed past either end is applied as typed rather than pulled back into range.
  3. Decide whether to prevent clipping. Left on, the gain is capped at whatever this particular file's peak allows, so a batch can end up with different files boosted by different amounts. Switched off, your figure is applied as typed and the clipped-sample count tells you what that cost.
  4. Pick what it is written back as. Match the input where possible, WAV at 16-bit, 24-bit or 32-bit float, or MP3 at one of seven bitrates from 64 to 320 kbps. The download keeps your filename with -volume added, and the summary reports the peak before and after.

Processing & privacy

Runs on your device. Your file is processed directly in this browser.

Common use cases

An interview recorded too quietly

A phone left on the table catches the conversation at a level nobody can hear through a laptop speaker. A lift of ten or twelve decibels makes the recording usable without arranging it again.

A music bed sitting under a voice

Cutting the track by a known figure, say −14 dB, before it reaches the edit is quicker and more repeatable than drawing a volume curve along the whole timeline.

A notification sound that startles people

Alert and ringtone files are usually mastered to full scale. Taking one down by 10 dB leaves it noticeable rather than alarming, and files that short are processed almost instantly.

Roughly levelling a folder of clips

Where recordings were made on different devices, running each through with its own figure gets them close to matching before anything is assembled.

Questions

Does my audio reach a server?
No. Decoding, the gain change and the re-encode all happen in this browser tab, so the samples themselves are never transmitted. The one thing that touches the network is the MP3 encoder, a script pulled from this site on your first MP3 run and reused after that; ask for a WAV and not even that is needed. Audio tools here that do send the file to a server say so on their own pages.
Why is the result not as loud as I asked for?
Prevent clipping measured the loudest sample and found that your figure would push it past full scale, so the gain was reduced until the peak sits just below the ceiling. The result line names both the gain applied and the one you requested.
How is this different from Normalize Audio?
Change Volume applies a relative change — the figure you type, up or down, whatever the file's current level happens to be. Normalize Audio measures the file first and works out the gain needed to reach a target you set, which is the better choice when several files have to end up matching.
Should I use percentage or decibels?
They set the same thing two ways. Percentage is amplitude, so 200% is twice the amplitude and roughly +6 dB; decibels is the unit other audio software uses, which makes it easier to reuse a figure taken from an editor or a mixing desk.
Why did my M4A come back as a WAV?
Browsers ship no AAC, FLAC or OGG encoder, so those formats can be read here but not written. The output is WAV, which is lossless but large, or MP3, which goes through the LAME encoder this page loads on demand and loses a little quality with each pass.
Do my ID3 tags and cover art survive?
No. The new file is written from the samples upward, so title, artist, album and artwork are all left behind. Audio Metadata can type the text tags back in afterwards, with two caveats: it writes ID3 into MP3 files only, so a WAV result is out of its reach, and it has no writer for cover art at all — artwork has to be re-attached somewhere else.
Will amplifying make a recording clearer?
It makes everything louder in the same proportion, hiss and mains hum included, so a noisy recording is no cleaner relative to the voice than it was. There is no noise reduction here — amplify a quiet recording and you amplify its noise floor with it.
Can I change the level of just one section?
Not in a single pass, because one factor is applied across the whole file. Trim Audio can cut the section out, this tool can change its level, and Merge Audio can put the pieces back together. Fade In & Out covers the start and the end.
Is there a length limit?
None is imposed, but the whole file is decoded into memory as floating-point samples, which costs roughly twenty megabytes per minute of 44.1 kHz stereo, and more than one copy exists while the work runs. Hour-long recordings make the tab work hard; a very long one can exhaust its memory.

Related tools