Volume Booster
Rescue a recording that came out way too quiet.
or paste one from your clipboard
How to use it
- Drop the quiet file onto the page. Before you touch anything, the tool measures the loudest sample in it and tells you how much headroom you have left.
- Pick a method: add a fixed number of decibels, or normalize so the loudest moment lands exactly on a level you choose.
- Press play. The transport plays the boosted version rather than the original, so what you hear is what you are about to download.
- If the clipping warning appears, lower the boost by the amount it names, or switch to normalizing. That method measures first and cannot clip.
- Choose an output format and download. The original file on your disk is never modified.
What a decibel actually adds
Decibels are a ratio, not an amount. Going up 6 dB doubles the amplitude of the waveform; going up 20 dB multiplies it by ten. Perceived loudness moves far more slowly than that. As a rough rule it takes about 10 dB before most people describe a sound as twice as loud. So a +3 dB nudge is a polite adjustment, +10 dB is a genuine change of character, and if you need +25 dB the original recording was very nearly silent.
Digital levels are quoted in dBFS, decibels relative to full scale, and they are always negative or zero. Zero is not a volume. It is the ceiling, the largest number the format can store. A file whose loudest sample reads −18 dBFS has 18 dB of headroom: you can add 18 dB before anything touches the ceiling, and every decibel past that starts flattening the tops of the waves.
Boosting versus normalizing
Adding a fixed amount is blind. You are telling the tool to multiply every sample by the same number, and whether that ends well depends entirely on what the file already contained. It is the right choice when you know how far off you are, or when several files need identical treatment so their levels stay in the same relationship to each other.
Normalizing measures first. The tool finds the single loudest sample, works out how far below your target it sits, and applies exactly that much gain and no more. It cannot clip, because the arithmetic is built around the peak. Its weakness is the mirror image: one stray click, a chair scrape, a door closing sets the level for the entire recording, and everything else ends up quieter than it needed to be. If a normalized file still sounds thin, find that one peak, trim or lower it, and normalize again.
When more gain is the wrong tool
Gain is indiscriminate. It lifts the speech and the air-conditioning hum by the same amount, so the ratio between them (the thing you actually care about) does not move at all. A recording that sounds quiet because it was made from across the room becomes a louder recording of a room, not a closer one. No amount of gain undoes distance.
The other case gain cannot solve is a file with a low average level but peaks already near the ceiling: a voice that drops to a mumble and then laughs. There is no headroom left to give it. That needs compression to bring the peaks down before any gain is useful. And if the goal is a delivery target rather than simply "louder", a podcast host, a streaming platform. Normalize to LUFS instead, which measures loudness across the whole programme rather than looking at one sample.