TRANSPOSE · SLOW DOWN · SPEED UP
Drop your audio file here or click to browse
MP3, WAV, FLAC, M4A, AAC, OGG, Opus · one file, up to 15 minutes
Every naive way of speeding a song up also raises its pitch, because on a tape, a record or a plain resampler the two are the same knob. Separating them takes a bit of work: the audio has to be stretched or squeezed in time first, in short overlapping pieces that are nudged until they line up, and only then resampled. Do both, with the amounts chosen so the unwanted half of each cancels the other, and you get a key change at constant speed or a speed change at constant key.
That is what this page does, on your own machine. Nothing is uploaded — the browser decodes the file, measures its tempo and key, does the processing and encodes the result. A four-minute track takes a few seconds, all of it spent locally. One thing worth knowing before you export a WAV into a DAW: the browser decodes everything at your sound card’s own sample rate, so a 44.1 kHz file normally comes back at 48 kHz.
Pitch is measured in semitones because that is how music is: twelve of them make an octave, and each one multiplies the frequency by about 1.0595. Speed is a plain percentage — 150% means the track finishes in two thirds of the time. The two ranges here stop at an octave and at half or double speed, and that is not an arbitrary limit. Past those points the grains being reassembled no longer resemble what the music was doing, and the result acquires a warble on held notes and a metallic edge on cymbals. Small moves are effectively transparent; a shift of three semitones or a slowdown to 80% will not be noticed by most listeners.
Tempo comes from onsets. The track is chopped into short frames, and every frame is compared with the last one to see where energy suddenly appeared across the spectrum — the signature of a drum hit, a plucked string or a hard consonant. That gives a spiky curve over time, and the spacing that repeats most strongly in it is the beat. The one thing this cannot settle by itself is whether a piece is at 75 or 150, since both fit the same spacing, so a candidate is scored together with its half and its double and the answer is reported in the range tempos are normally quoted in.
Key comes from the balance of the twelve notes. The track is turned into a running tally of how much of each pitch class it contains, and that tally is compared against a profile of each of the 24 major and minor keys, drawn from studies of what listeners hear as belonging in a key. The best fit wins, but the number reported as confidence is the margin over the runner-up rather than the winner's own score — because a track that fits C major at 0.90 and A minor at 0.89 has not really been identified, and it would be dishonest to say so.
Yes, and that is the default. The two sliders are separate: move pitch and the song plays in a different key at the same speed, move tempo and it plays at a different speed in the same key. Speeding a file up the naive way raises its pitch as a side effect, because both come from the same clock — this tool stretches the audio in time first and then resamples it, and the two operations are chosen so the unwanted half of each cancels out.
Use the presets. Nightcore raises the pitch four semitones and runs the track at 125%. Slowed drops it two semitones and runs at 85%. Both are just starting points — move either slider afterwards and the preset stops applying. For the classic sped-up sound where pitch and speed rise together, tick Move them together, which reproduces what playing a record too fast does.
On music with an audible beat it is normally within a beat per minute of the true tempo. It works by finding where energy suddenly appears across the spectrum — which is what a drum hit looks like — and then finding the spacing that repeats. Two things can throw it off: music with no percussion gives it very little to lock on to, and a track whose beat is ambiguous can come back at half or double the tempo you would count, because 75 and 150 genuinely fit the same evidence. The confidence figure next to the reading tells you which case you are in.
Treat it as a strong suggestion rather than a verdict. It measures how much of each of the twelve pitch classes the track contains and compares that against the profile of every major and minor key. That works well on tonal music that stays in one key, and less well on a track that modulates, on something heavily atonal, or on a piece where the relative major and minor are equally plausible — C major and A minor use the same seven notes, and only the weighting separates them. The confidence figure is the margin over the runner-up, so a low number means two keys fit almost equally.
No. The file is decoded, analysed, processed and encoded by your own browser, and nothing is sent to a server at any point. You can prove it: load the page, disconnect from the internet, and the tool still works end to end. That is also why long files take a moment — the work is being done on your machine rather than on ours.
Because there is no way to invent the detail a big shift needs. The audio is cut into short overlapping grains that are nudged to line up before being reassembled, and the further you move from the original the less each grain resembles what should be there — which is heard as a slight warble on sustained notes and a metallic edge on cymbals. Up to about three semitones it is very hard to hear. Beyond seven it is obvious on almost any material, which is why the range stops at an octave rather than pretending further is useful.