Slowed + Reverb Maker

It starts the way it always does: a track you have heard a hundred times shows up in your feed at three-quarter speed, wrapped in enough echo to fill a parking garage, and suddenly the song you were tired of is the only thing you want to hear at 1 a.m. The vocals sit lower. The drums sprawl. Someone did something to this song, and it took them less time than finding their headphones.

So you look it up, and the results are a swamp: Discord servers with invite links that expired in 2021, desktop DAWs with a 4 GB download, and ten different 'slowed + reverb' websites that upload your file to a server you know nothing about and hand back an MP3 with a watermark. You wanted to slow one song down, not install a lifestyle.

This page is the sane middle path. Drop in an audio file you have the rights to, drag two sliders — speed and reverb — and render. Everything happens in your browser: decoding, slowing, the reverb itself, the export. Your file never uploads anywhere, which matters when the track is unreleased, licensed, or just nobody's business but yours.

A slowed + reverb maker applies two transformations to an audio file and hands back the result. Transformation one: playback speed below 1.0×, which stretches duration and drops pitch together — the exact behavior of grabbing a record and slowing the turntable. Transformation two: artificial reverb, a synthetic sense of space added on top, usually a big one, so the already-drowsy sound gets a halo.

The two effects earn their pairing. Slow alone sounds like a broken tape deck; reverb alone is a band in a gymnasium. Together, the speed drop gives weight and the reverb gives atmosphere, and the combination hits a specific emotional register — nostalgia, late drives, the feeling of a memory you are not quite done having. That is why the style survived from Houston basement tapes to a billion TikTok sounds.

A browser-based version adds the properties you actually want from a ten-second creative job: no install, no account, no upload, no watermark. Open the tab, drop the file, move the sliders, render, download. Close the tab and nothing remains anywhere.

Credit where it is due: the aesthetic's grandfather is DJ Screw, Robert Earl Davis Jr., who built his reputation in early-1990s Houston by replaying hip-hop records drastically slowed — 60 to 70 percent of original speed — with custom beat repeats and deep-voiced local MCs on top. He called it chopped and screwed, sold it as cassette tapes from his house, and the sound became a regional dialect before it became a national one. He died in 2000; the idea did not.

Two decades later the recipe simplified. Producers on SoundCloud in the mid-2010s realized you could skip the chopping entirely: slow a track to about 0.85×, drown it in free convolution reverb, and the emotional payload arrived intact. The format was born for the platform — long ambient uploads for late-night listening, reposted in chains. When TikTok's music library met the style, 'slowed + reverb' edits became a lazy shorthand for longing, and suddenly every pop hit had a slower, sadder twin competing with it.

The family tree has other branches worth knowing. Nightcore is the same operation with the sign flipped — speed up to 1.15× or 1.25×, pitch rises, energy replaces melancholy. Vaporwave slowed things down too, but added tape wobble and nostalgia samples. And the purest ancestor of all is the turntable itself: a 45 rpm record played at 33⅓ runs at 0.7407× speed, which is why half the slowed tracks in the world land suspiciously near that ratio. The mathematics of the format is older than the people making it.

There are two ways to make audio slower, and the slowed+reverb style is built on the older, simpler one. Varispeed changes the playback rate itself: every sample gets read more slowly than it was written, so duration stretches and frequency shrinks by the same factor. Double the duration and pitch falls one octave. The music stays internally perfect — all intervals, all ratios — just transposed down and stretched out, exactly like slowing a record or a tape.

The other way is time-stretching, which attempts to slow duration while holding pitch — the algorithm cuts the audio into overlapping grains, spreads them apart, and stitches the seams. Done well it is invisible; done cheaply it smears transients into mush. Crucially, even done well it is not the sound people mean when they say slowed. The lowered voices are the point. The genre's whole emotional mechanism is the pitch drop.

The arithmetic is clean. Pitch shift in semitones equals 12 × log₂(speed). At 0.90×: −1.82 semitones. At 0.85×: −2.81, a touch under a minor third — the canonical slowed sound. At 0.80×: −3.86, nearly a major third. At the vinyl ratio of 0.7407×: −5.36, a fat perfect fourth. Durations scale by the reciprocal: a 3:30 song at 0.85× becomes 3:30 ÷ 0.85 = 4:07. Slower is heavier, and heavier is only better until the vocal turns to syrup — past about 0.75×, words start dissolving.

Reverb is the sound of a room. Clap once in a stairwell and you do not hear one clap — you hear the clap, then hundreds of tiny echoes off every surface, arriving at slightly different times and blending into a shimmer. Record that stairwell's response to a clap and you have an impulse response: a file that mathematically describes how that specific space colors and decays sound. Feed any audio through that description and the audio sounds like it happened in the stairwell. That process is convolution, and it is how essentially all realistic digital reverb works.

The knobs translate simply. Decay time is how long the echo takes to fade to nothing — 1 second feels like a living room, 3 seconds like a concert hall, 8 seconds like a cathedral and, in this genre, often too much. Wet/dry mix is the ratio of echo-soaked signal to untouched signal: 0% is bone dry, 100% is only the room with the performance drowned in it, and 30–40% wet is the genre's sweet spot. Pre-delay is a short silence between the dry sound and the reverb's arrival — even 20 milliseconds of it keeps lyrics intelligible, because your ear needs the clean syllable first to lock onto the words.

This tool builds its spaces mathematically rather than sampling real ones: an impulse response synthesized as random noise multiplied by an exponential decay, shaped to the decay time you choose. Noise-in-a-decay-envelope is the classic artificial reverb recipe — mathematically it behaves like a diffuse, characterless 'big room,' which is exactly what the style wants. A real concert-hall IR brings its own personality; the synthetic one brings only size, staying politely out of the song's way.

The workflow is deliberately short — the genre's whole charm is that the transformation is one gesture, not a session. Six steps, most of them optional.

1. 1

   Drop in your audio file

   MP3, WAV, M4A, FLAC, or WebM — the browser's decoder handles it locally, and the file never uploads. Only use recordings you have the rights to remix.

2. 2

   Set the speed

   Start at 0.85×. The duration readout updates live so you can see exactly how long the edit will run. Push toward 0.80× for the heavier night-drive feel.

3. 3

   Choose the reverb size

   Small room, hall, or cathedral — decay time from 1 to 8 seconds. Hall at 2.5–4 seconds is the default sound of the genre.

4. 4

   Balance wet and dry

   Begin at 35% wet. Preview a chorus with vocals and adjust: muddiness means less wet, sterility means more. Pre-delay at 30 ms if words blur.

5. 5

   Preview before rendering

   Scrub the timeline and audition the loudest section — drums and chorus expose over-wet settings faster than a quiet verse does.

6. 6

   Render and download

   The offline render processes faster than real time and automatically pads the end so the reverb tail decays fully instead of clipping off. Download appears when it finishes; nothing is stored anywhere after you close the tab.

Real-time playback and offline rendering are different jobs, and this tool uses the right one for each. While you preview, the audio graph runs live: your decoded file plays through an AudioBufferSourceNode with its playbackRate set to your chosen speed — one property change producing both the stretch and the pitch drop, because varispeed is the native behavior of sample playback. The signal then forks through a dry gain path and a wet path containing a ConvolverNode loaded with the synthesized impulse response, and the two paths sum at the output. That is the entire signal chain: rate, convolve, mix.

For export, the same graph is rebuilt inside an OfflineAudioContext — a Web Audio construct that renders the chain as fast as the CPU can grind, with no output device involved, writing samples straight into an output buffer. A three-minute song commonly renders in a handful of seconds. The render length is deliberately set to song length plus the decay time, because convolution's output legitimately extends past the input's end: the last note's echoes have to land somewhere, and chopping them at the final sample is the tell of a careless implementation.

The impulse response itself is generated per your settings: unit-variance noise across the buffer length, multiplied by exp(−6.9 × t / decayTime) — the envelope that falls 60 decibels over the chosen decay time, which is the textbook definition of T60. A short stereo decorrelation between left and right channels keeps the result from collapsing into mono. Convolution then happens inside the browser's highly optimized DSP, the same math plugin reverbs have used since the 1990s, just wrapped in a page instead of a product authorization.

Three edits, computed honestly. Durations use the reciprocal of speed; pitch shifts use 12 × log₂(speed); tail padding equals the decay time.

A rule hiding in those numbers: the slower you go, the drier you should run. Heavy slow already smears diction through the pitch drop; stacking a cathedral on top removes what is left. The genre's best-known edits sit near 0.85× with moderate wet precisely because that is where vocals survive. Break the rule on purpose or not at all.

The presets below cover the practical range. Anything below 0.70× is performance art; anything above 0.95× barely reads as slowed.

The final row is the traditionalist's choice and a useful calibration: 33⅓ divided by 45 is 0.7407, so any record person can identify it blind. If your edit at that speed sounds right, your monitoring is honest; if it sounds muddy, fix the mix, not the speed.

Half the outcome is decided before you touch a slider. Slowing magnifies whatever the mix already contains: a sharp, well-produced track becomes dreamlike; a loose, muddy mix becomes soup. The edits that go viral almost always start from recordings with clean vocal presence up front, punchy but simple drums, and arrangement space — because when everything drops nearly three semitones, whatever was loudest stays loudest, and whatever was cluttered gets twice as cluttered in feel.

Temperament matters as much as production. Major-key, mid-tempo songs with bright toplines take best — the drop into minor-third territory adds a bittersweet gravity that the original never had. Tracks that are already slow, already saturated with natural room sound, or already melancholy tend to collapse: you cannot make a dirge more wistful by stretching it. Mid-2010s synth-pop, disco-adjacent dance records, and anything with a soaring hook are the classic fodder, which is exactly what the trend's biggest edits chose.

Mastering counts double. Loud, brick-walled masters keep their density after slowing, so the effect reads as heavy rather than spacious; a track with dynamic range leaves room for the reverb to bloom into. If your file is a crushed festival master, run the wet mix lower than the table suggests and lean on pre-delay to carve space back out. And always audition the chorus, not the intro — intros flatter every setting, and the chorus is where wet levels either sing or drown.

• Best raw material: bright vocal-forward mid-tempo pop, 100–130 BPM originals
• Surprisingly good: acoustic singer-songwriter recordings with honest dynamics
• Risky: dense trap or metal mixes — slowing thickens walls that were already walls
• Usually wasted: ballads, ambient pieces, anything under 80 BPM at the start

The tail-clip trap is the classic broken export: a render whose length equals the song's length slices the reverb off at the last sample, ending in an audible snap of silence. This tool pads the render by the decay time automatically, but if you rebuild this effect in a DAW, that padding is your job — add three to five seconds to the timeline and let the fade land.

Very long files test memory, not CPU: the decoded audio, the convolver, and the output buffer all live in RAM simultaneously, so a 90-minute FLAC on a phone with a dozen tabs open may fail politely. Render long-form audio on the machine with headroom. Format-wise, the export inherits what the browser's encoder supports — WebM/Opus in most browsers, MP4/AAC from Safari — and any converter can transpose if a platform insists on MP3.

And the limit that is not technical: copyright. A slowed version is a derivative work of the original recording — the pitch drop does not launder the rights. Uploading someone else's song, however transformed, invites takedowns, muted videos, and demonetization, and 'everyone on TikTok does it' has never once worked as a legal position. Use your own recordings, properly licensed stems, or cleared material. The effect is free; the music is not.

The obvious crowd: fans building mood edits of tracks they love, and TikTok creators hunting the sound that makes a clip feel like 2 a.m. Producers use it as a quick arrangement audition — hearing a beat at 0.85× reveals which elements carry and which collapse, and some genres were arguably discovered exactly that way. Podcasters and video editors slow music beds down twenty percent to stretch a cue under a montage without an audible edit point.

Then there are the sleep and study listeners, who drag their own playlists down a few percent and add soft rooms because the result sits under concentration without demanding attention. Musicians practice to slowed versions of difficult passages — the varispeed pitch drop matters less than the extra time between notes. And every so often someone drops in a voice memo of a family member and walks out with something that feels like a memory. That one is worth the whole tool.

If you are layering voice over your edit — an intro, a spoken-word bed — the Online Voice Recorder captures clean local audio for it. And when the remix needs a looping visual to match, Video to GIF turns a clip into a shareable loop without leaving the browser either.