Fade Audio In and Out

A recording that starts or stops at a non-zero sample produces a click, because the speaker cone is being asked to jump instantly from one position to another. A fade removes it by moving the level gradually instead, and the length needed is often far shorter than people expect — a twentieth of a second is enough to make an edit inaudible while still sounding like a cut rather than a fade.

• Podcast intros and outros. Half a second at each end removes the room tone snapping on and off.
• Music outros. Five to ten seconds is the range almost every commercial fade sits in.
• Edit points. A tenth of a second either side of a cut hides the join.
• Background beds. A long fade under a voiceover, or a fade down to a level rather than to nothing.

A fade curve is the shape the level takes on its way down. All six here reach true silence, which sounds obvious and is not — several of the shapes commonly offered elsewhere stop short of it and leave an audible step at the exact moment the track is supposed to have ended.

• Linear. A constant rate of change, half level at the halfway point. The safe default.
• Equal power. Holds perceived loudness steady, -3 dB at the midpoint. Use it wherever two sounds overlap.
• Smooth. Eased at both ends, so neither the beginning nor the end of the fade is audible as an edge.
• Exponential. Holds its level and then falls away — an ending that feels deliberate.
• Logarithmic. Drops immediately and leaves a long quiet tail.
• Steep. Nearly a hard stop, without the click one produces.

Each fade carries its own position, so a fade in can begin partway into a recording and a fade out can finish before the file does. Moving a fade off the file's edge raises a second question, and the tool asks it rather than deciding for you: what happens to the rest.

• Trim it — the file ends where the fade ends. This is what "end the track at 2:30" almost always means.
• Leave it — the remainder stays as silence, which is what you want when the fade sits inside a longer bed.
• Either way, the resulting length is shown beside the controls before anything is rendered.

A fade does not have to end at nothing. Set a floor and the level lands there instead — a music bed ducking under a voiceover, an ambience track pulled back for dialogue, or an intro that settles rather than disappears. The floor runs from -1 dB down to -60 dB; below that there is nothing left to hear, which is why silence is its own setting rather than the bottom of the range.

A track that has to repeat — an ambience bed, a game loop, a podcast underscore — clicks at the repeat point unless the end and the beginning happen to line up, which they almost never do. The seamless loop mode fixes it by crossfading the end of the track over the beginning, so the moment the file repeats is a continuation of the audio rather than a join.

• The output is shorter. By exactly the length of the overlap you chose, and the tool shows the new length before it renders.
• There is no cut to hide. Both joins are sample-exact continuations of the source, so nothing is being disguised.
• Equal power is the default here. It keeps the perceived loudness steady through the overlap, which is what stops a crossfade dipping in the middle.

The length of your file is read first, because everything else depends on it: how long a fade can be, how far it can move and how much overlap a loop can take are all bounded by the audio you actually have. The fade is then a per-sample multiplication — the level at each moment times the curve at that moment — and because that is exactly what the picture on screen draws, what you see before you press the button is arithmetically the file you get.

• Runs on your device, not our servers. The file is read, faded and written where it already is.
• The finished length is checked. Against the length the tool planned, before the result is handed back — a fade that changed the duration when it should not have is a fault, not a result.
• Asking for no change costs nothing. With both fades at zero and the format unchanged, the audio is copied rather than re-encoded, so no quality is lost to a round trip you did not want.

Powered by optimised audio processing algorithms.

• 0.05-0.2 s — de-clicking an edit point. Short enough that it reads as a cut.
• 0.5-1 s — podcast and voiceover intros. Removes the room tone snapping on.
• 3-5 s — a song ending that should feel like an ending.
• 8-20 s — a long cinematic fade, or a bed sinking under narration.
• Match the fade to the music, not the clock. A fade that ends on a beat sounds intentional; one that ends between beats sounds like a mistake.

• Audio Cutter
• Audio Merger
• Volume Booster
• Audio Normalizer
• Audio Converter
• Ringtone Maker