Join several audio files into one continuous track. Drop them in, put them in order, choose a silent gap or an overlap between them, and export a single WAV. The crossfade is equal-power rather than linear, which matters: two linear ramps dip about three decibels where they meet, and that dip is audible as a dropout at every join. A mono clip joined onto a stereo one is widened rather than the stereo being flattened. If the files have different sample rates you are told, rather than getting a result where one clip plays nine per cent fast. Nothing is uploaded.
| Option | Effect on length | Best for |
|---|---|---|
| Butt join | Sum of the clips | Already-trimmed material |
| Gap | Longer by the gap | Spoken segments |
| Crossfade | Shorter by the overlap | Music and ambience |
| Normalise | None | Clips at different levels |
A gap inserts silence and makes the file longer. A crossfade overlaps the clips so one fades out as the next fades in, making the file shorter by the overlap. Use a gap between spoken segments and a crossfade between musical ones.
Because two linear ramps summed at the join dip by about three decibels in the middle, which you hear as a dropout every time. A cosine-and-sine pair sums to constant power, so the level holds steady across the transition.
You are told, and the merge stops. Concatenating raw samples at different rates makes one clip play at the wrong speed — a 44.1k clip in a 48k file runs about nine per cent fast and sounds slightly sharp. Convert them to a common rate first; resampling is a quality decision that should be yours.
The mono one is widened to match, so it plays on both sides. Doing it the other way would silently throw away half of the stereo clip.
Scales everything so the loudest sample sits just under full scale. It is peak normalisation, not loudness normalisation — it will not make a quiet clip and a loud clip sound equally loud, particularly if the quiet one has one loud moment in it. For matching perceived loudness you need a proper loudness tool.
Only your browser's memory, since everything is decoded at once. Long files are large uncompressed: a minute of CD-quality stereo is about 10MB.
No. Everything is decoded and joined in your browser.
Each file is decoded to raw samples, then joined by writing them one after another into a single output allocated once at the final length, rather than by repeatedly concatenating arrays. Three details decide whether the result sounds right. The crossfade is equal-power — cosine out, sine in — rather than linear. Two linear ramps summed across a join dip by about three decibels in the middle, which is audible as a dropout at every transition; the cosine and sine pair sums to constant power instead, which is what a crossfade is for. Channel counts are reconciled upward: joining a mono clip onto a stereo one widens the mono rather than truncating the stereo, because the obvious direction silently discards half of the other clip. And sample rates are checked rather than assumed — clips at different rates concatenated as raw samples play at the wrong speed, a 44.1kHz clip inside a 48kHz file running about nine per cent fast and slightly sharp, which sounds like a mistake rather than an error. Rather than resampling silently, mismatches are refused with an explanation, because resampling is a decision about quality that belongs to the person making the file. Optional peak normalisation scales everything so the loudest sample reaches just under full scale, and the page is explicit that this is peak rather than loudness normalisation: it makes a quiet clip reach the same ceiling as a loud one, which is not the same as making them sound equally loud, and a quiet recording with one door slam in it gets almost no lift. Export is WAV at 16, 24 or 32-bit float, with 32-bit written as format tag 3 so it is not misread as integer PCM. Nothing is uploaded.