How long should a crossfade be when you join audio?
Short answer: the length does not change how deep the blend dips, only how long it lasts. In our test of the audio joiner, every crossfade from 0.5 s to 4 s left two different tracks about 3 dB quieter in the middle of the overlap than either track alone. What changed was the time spent more than 1 dB down: 0.38 s with a 0.5 s crossfade, 1.53 s with 2 s, 3.07 s with 4 s.
- Pick the length by how long you want both tracks audible at once. Short (0.5 to 1 s) changes quickly; longer settings blend slowly. Each seam also shortens the joined file by the crossfade length.
- Leave Fade in and Fade out at 0 on a crossfaded seam. With a 2 s fade-out, a 2 s fade-in and a 2 s crossfade on the same seam, the middle dropped 9.1 dB instead of 3.
- The crossfade is capped at half of the shorter neighbouring clip. On two 8 s clips, setting 5 s gave a 4.0 s overlap while the slider still read 5.0 s. Check the total length instead.
The setting is the "Crossfade between clips" slider in the audio joiner, from 0 to 5 s in 0.1 s steps. One setting applies to every seam.
Why a crossfade dips in the middle
During a crossfade the first clip's volume falls in a straight line from full to zero while the second clip's rises from zero to full, and the two are added together. In the joiner the two volumes always sum to 1, so halfway through each clip plays at half volume (-6 dB).
What you hear then depends on how alike the two clips are at that moment:
- Two unrelated passages (two different songs, or two different parts of one song): their waveforms do not line up, so their powers add rather than their amplitudes. Two halves in volume are two quarters in power, half the power in total, which is 3 dB down.
- The same sound in step (the same sustained note, lined up): the amplitudes add, 0.5 + 0.5 = 1, and there is no dip.
- The same sound out of step: the two cancel, and halfway through the formula gives silence.
A blend that keeps the power constant would need curved volume ramps instead of straight ones. The joiner on this site uses straight ones, so the 3 dB figure below is what you get with it.
How we tested
- Tool: this site's audio joiner, run in Chromium 153 through Playwright 1.63 on 2026-10-01. Two clips per job, crossfade slider set, WAV export. For every job we also rebuilt the join offline from the joiner's own formula; all 18 exports matched it to within 1 step of 16-bit (1 LSB).
- Two different tracks at the same level: two independent 8 s white-noise clips, stereo, both at -20 dBFS. Noise has a steady level, so any dip is the crossfade's doing.
- Music: two 10 s passages (1:00 to 1:10 and 1:15 to 1:25) of Mozart's Symphony No. 40, first movement, played by the Musopen Symphony, public domain on Wikimedia Commons, converted to 44.1 kHz 16-bit stereo.
- Same sound: two 8 s 420 Hz sine tones at -12 dBFS, once in step and once with the second one inverted.
- Measure: level in 100 ms windows every 10 ms. We compared each window with the level a constant-power blend of the same two clips would have at that point, so a passage that is louder or quieter on its own does not count as a dip. For two clips of equal level that is simply the level before the seam.
Results: the dip is the same at every length
| Crossfade | Level in the middle | Time more than 1 dB down | Time more than 2 dB down | Joined length (from 2 × 8 s) |
|---|---|---|---|---|
| 0 (none) | no dip | 0 | 0 | 16.0 s |
| 0.5 s | -3.0 dB | 0.38 s | 0.25 s | 15.5 s |
| 1 s | -2.9 dB | 0.77 s | 0.51 s | 15.0 s |
| 2 s | -3.1 dB | 1.53 s | 0.99 s | 14.0 s |
| 3 s | -3.1 dB | 2.32 s | 1.51 s | 13.0 s |
| 4 s | -3.0 dB | 3.07 s | 2.02 s | 12.0 s |
The time spent more than 1 dB down is about three quarters of the crossfade every time, and the time more than 2 dB down about half. That is what the straight-line formula predicts for two unrelated clips: -3.01 dB at the centre, 76.7 % of the overlap below -1 dB and 51.2 % below -2 dB.
| Crossfade, Mozart | Level in the middle | Lowest point | Time more than 1 dB down |
|---|---|---|---|
| 0.5 s | -2.9 dB | -3.6 dB | 0.36 s |
| 1 s | -3.0 dB | -5.1 dB | 0.78 s |
| 2 s | -3.0 dB | -4.1 dB | 1.59 s |
| 4 s | -3.2 dB | -5.0 dB | 3.32 s |
Real music gave the same middle figure. The lowest point fell further, 3.6 to 5.1 dB, where the noise clips never went past 3.1 dB: an orchestra's level moves from one 100 ms window to the next, and steady noise does not.
Settings that make the dip deeper
| Seam settings, two 8 s noise clips | Level in the middle | Time more than 6 dB down | Joined length |
|---|---|---|---|
| Crossfade 2 s | -3.1 dB | 0 | 14.0 s |
| Crossfade 2 s + Fade out 2 s on clip 1 + Fade in 2 s on clip 2 | -9.1 dB | 0.81 s | 14.0 s |
| No crossfade, Fade out 2 s + Fade in 2 s | quietest window -56.9 dBFS (from -20) | 2.01 s | 16.0 s |
| Crossfade 2 s, second clip 6 dB louder | -3.0 dB; quietest window 1.1 dB below the first clip | 0 | 14.0 s |
| Crossfade 2 s, same 420 Hz tone in step | 0 dB | 0 | 14.0 s |
| Crossfade 2 s, same tone, second one inverted | -30.8 dB (100 ms window) | 1.01 s | 14.0 s |
The per-clip fades and the crossfade multiply. With all three set to 2 s on one seam, each clip is at a quarter of its volume halfway through instead of half, so the middle lands about 9 dB down. Fades without a crossfade do not overlap at all: the first clip fades to silence, then the second fades up from it, which is a pause, not a blend.
When the second track is louder, the blend never falls far below the quieter one; with a 6 dB difference the lowest window was 1.1 dB under the first clip's level.
The tone rows are the extreme case of two clips that hold the same note. In step there is no dip at all; inverted, the two cancel through the middle of the overlap. Most music is neither, but two loops or two takes of the same sustained chord can land anywhere between, so preview such a seam before you export.
The 5-second setting is not always 5 seconds
The joiner limits each crossfade to half of the shorter of the two clips at that seam, so a crossfade can never swallow a whole clip. On two 8 s clips the 5 s setting produced a 4.0 s overlap and a 12.0 s file, the same as the 4 s setting. On two 3 s clips it produced 1.5 s and a 4.5 s file. The slider kept showing 5.0 s in both cases; the total length the joiner shows did change, to 0:12.000 and 0:04.500.
What to set
- Different songs: choose the length by how long you want to hear both. The middle dips about 3 dB whatever you choose; a short crossfade keeps that dip short.
- Leave Fade in and Fade out at 0 on the clips at a crossfaded seam. Use the per-clip fades for the very start and the very end of the joined file.
- Parts of one continuous recording: use crossfade 0. A crossfade overlaps and removes music at every seam. If MP3 parts leave a short silence at each seam, see why there are gaps between joined MP3 files.
- Short clips: keep the crossfade under half of the shortest clip, or it will be shortened for you. Read the total length to check.
- Preview the seam with the play button before exporting, and export WAV if you will edit the result again.
Questions
Can I set a different crossfade for each seam?
No. The joiner has one crossfade slider and applies it to every seam, except where a clip is too short and the cap above applies.
Does the preview match the exported file?
The preview and the export are built by the same function in the page, so the crossfade is identical. We measured the WAV exports; an MP3 export adds its own encoding on top.
Is a 3 dB dip audible?
We measured levels, not listeners, so we cannot give a figure for when people notice it. What the numbers do say is that it is the same depth at every length, and that stacking fades on the seam makes it three times deeper in decibels.