Slowed + reverb. Without uploading the song.

What the slider actually does

Slowed and reverb is not a "slow down" effect in the way a DAW means it. It is the sound of a record played at the wrong speed: the tempo drops and the pitch drops with it, locked together. That coupling is the entire genre. Pull a track to 0.82× and everything in it — the vocal, the snare, the bassline — arrives 3.4 semitones lower, which is most of a minor third. A three-and-a-half minute song becomes four minutes sixteen.

The table below is the exact relationship, because it is not a matter of taste. The pitch shift in semitones is twelve times the base-two logarithm of the speed, so 0.75× is almost exactly five semitones down — a fourth — and 0.70× is a hair past a tritone.

If you want the tempo to drop and the pitch to stay where it is, this is the wrong tool and the result will not sound like the genre — use change audio speed, which preserves pitch with a time-stretch. If you want the pitch to move and the tempo to stay, that is the pitch and key changer. This page is deliberately the one that moves both, because that is the effect people are actually asking for.

The reverb, and why it is not just "more"

The reverb here is convolution reverb: the track is convolved with an impulse response, a recording of how a space answers a single sharp sound. The impulse is synthesised rather than sampled from a real hall, which keeps the page free of a download and makes the character adjustable, and it is built from decaying noise with three properties that matter more than the wet level.

The two channels are independent noise. Using the same noise for left and right produces a tail that collapses to the centre of the stereo image — technically reverb, but it sits on top of the track like a mono wash rather than around it. Independent noise per channel measures near zero correlation between the two sides of the tail, which is what makes it sound like a space rather than an effect.

There is a 30-millisecond pre-delay. In a real room the direct sound reaches you before the first reflection does. Without that gap the wet signal smears straight over the dry one and the vocal loses its edge; the result sounds washed out rather than spacious. Thirty milliseconds is short enough that you do not hear it as an echo and long enough that the vocal keeps its consonants.

The wet path is low-passed at 3.8 kHz. Air and soft surfaces absorb treble faster than bass, so a real tail gets darker as it decays. An undamped tail keeps all its top end and sounds like a metal tank. This is the single setting most responsible for whether a slowed edit sounds warm or cheap.

The dry signal is left at full level and the wet is added on top, rather than crossfading between them. Crossfading is how a mixing desk thinks about a send, but it is wrong here: as you add reverb the vocal would recede, and the genre wants the vocal where it is with a room behind it. The consequence is that the output peak goes above the input peak, so the render is pulled back down when it would clip and left alone when it would not. On a commercially mastered file — which is everything you are likely to drop here — the standard setting measured a peak of 1.14 before that pull-back, and the heaviest settings 1.23. Both would have clipped.

What you lose, and what to do about it

Slowing a track exposes whatever the encoder threw away. An MP3 at 128 kbps already has a lid on it somewhere around 16 kHz, and playing it at 0.82× drags that lid down to about 13 kHz where you can hear it as a dullness in the cymbals. If you have a lossless copy, use it: drop the FLAC or WAV in here and choose WAV output, and the only generation of compression is the one you add at the end.

Editing an MP3 always costs one compression generation, because the audio has to be decoded and re-encoded. At 192 kbps and above that is not something most people can hear on a slowed edit, where the top end is already down. At 128 it stacks with the loss above. The bitrate control is there rather than fixed for that reason.

Low end is the other thing to watch. Dropping 3.4 semitones moves a bassline that sat at 60 Hz down to about 49 Hz, and moves a kick fundamental below what a phone speaker will reproduce at all. The track will sound thinner on a phone and heavier on headphones than the original did. If it comes out muddy rather than heavy, the fix is usually a small cut in the low mids rather than less reverb — the EQ and bass tool is next door.

Vocal intelligibility falls off a cliff somewhere below 0.75×. Consonants are short, high-frequency events, and stretching them while damping the top end takes away both of the things that make them legible. That is a feature at 0.70×, where nobody is listening for the words, and a problem at 0.75× if the song is wordy.

Practical notes

Working from a full song. If you want the slowed treatment on the vocal only, or an instrumental to slow without the vocal smearing, run the track through the stem splitter first and bring the part you want back here. That is neural source separation running on your own machine, and it is the usual first step for a serious edit rather than a quick one.

Length changes, so tags and playlists shift. The output is a new file with a new duration. Nothing carries the original's title and artist across, so if you are building a library, the tag editor writes those back without re-encoding the audio a second time.

Batch works. Drop a folder's worth and every file gets the same treatment, returned as a zip. The render is fast enough that an album is a matter of a minute or two; the slow part is the encode, not the effect.

On uploading the result. A slowed edit of someone else's recording is still their recording. Platforms match it — pitch and speed changes do not defeat content identification, and have not for years — so expect it to be claimed or muted if you post it publicly. Making one for yourself is a different matter from monetising one, and this page has no opinion about which you are doing; it is worth knowing which rules apply before you upload.

It works offline. Nothing here needs the network after the page has loaded — the effect is Web Audio, which is built into the browser, and no codec is downloaded unless you feed it a format that needs one. Turn the wifi off and it still runs.