Reversing the Room: What Pre-Verb Actually Does
Data engineer who loves building high-performance data and web-related tools. Creator of SlowedReverbMaker.net, implementing browser-side digital signal processing (DSP) to democratize audio editing.
You have heard this effect thousands of times without necessarily having a name for it. A vocal line that seems to inhale before the first word arrives. A snare preceded by a swelling rush that collapses exactly on the hit. A transition into a chorus where the energy gathers out of nothing.
That is reverse reverb, or pre-verb: reverberation that arrives before the sound that caused it rather than after. It is physically impossible, which is exactly why it works as a signal that something is about to happen.
The standard way to make it involves four steps and a lot of file management. There is a much shorter route, and it comes from noticing that the four steps are algebraically equivalent to one.
Try this while you read: Reverse Reverb
Make reverb swell into a sound instead of after it. Free, in your browser, no signup.
1. What Reverb Is, Mathematically
To see the shortcut you need one idea, and it is worth stating precisely because everything else follows from it.
- Impulse response
- A recording of how a space answers a single instantaneous burst of sound. It starts loud — direct sound plus earliest reflections — then decays as energy is absorbed by surfaces.
- Convolution
- For every sample in your audio, stamp down a scaled copy of the entire impulse response starting at that point, and add all the copies together. Each sound gets its own decaying tail, and the overlapping tails become the reverb.
- Pre-verb
- Reverberation that arrives before the sound that caused it rather than after. Physically impossible, which is exactly why it works as a signal that something is about to happen.
The Property That Matters
An impulse response fully describes a space's linear behaviour. To make any other sound appear to have been played in that room, you convolve it with the impulse response.
So a reverb is completely characterised by one recording. Change the impulse response and you change the room — and that is the lever this whole technique pulls on.
2. The Recipe Everyone Uses
The traditional way to produce pre-verb in a DAW is a four-step bounce. It long predates convolution plugins — it was originally done by physically flipping tape.
- Reverse the audio you want the effect on.
- Apply reverb to the reversed audio.
- Render, or bounce, the result to a new file.
- Reverse that file back.
Why It Works
While reversed, the reverb decays forward in the normal way. When you flip the result back, everything reverses — including the tails — so each decay becomes a swell that builds into its sound instead of trailing away from it.
What It Costs
You cannot hear the result while adjusting it, because the effect only exists after the second reversal. So it becomes guess, render, reverse, listen, undo everything, guess again.
Every change to the reverb means repeating all four steps, and you accumulate intermediate files you have to keep track of.
3. The Identity That Removes All Four Steps
Here is the observation that collapses the whole procedure into a single render.
The four steps are: reverse the audio, convolve with the impulse response, then reverse the result. Written compactly, that is reverse(convolve(reverse(x), h)) — where x is your audio and h is the room.
The Algebra
Convolution has a property that makes this collapse: reversing both inputs reverses the output. Consequently, reversing the audio, convolving, and reversing back is exactly the same as leaving the audio alone and convolving it with a reversed impulse response:
reverse(convolve(reverse(x), h)) = convolve(x, reverse(h))
The left-hand side is the four-step bounce. The right-hand side is a single convolution in which the only thing that changed is the room. You never touch the audio at all — you reverse the impulse response, which is typically a fraction of a second long, and use it as a normal reverb.
This is not an approximation or a close-enough substitute. It is an identity, and the two routes produce the same samples.
Verifying It Rather Than Trusting It
Claims like this are worth checking, so I implemented both routes directly and compared them sample by sample.
What This Means for the Tool
This is what the reverse reverb tool does. It reverses the impulse response and renders once. There is no bounce, nothing to re-import, and because it is a single signal path rather than a two-stage process, what you hear while adjusting the controls is by construction the same thing that exports.
4. The Complication Nobody Mentions
The identity is clean, but implementing it naively produces a subtly wrong result — and it took measuring to see why.
The Swell Lands in the Wrong Place
A reversed impulse response swells rather than decays, but it is still positioned after the sound that triggered it. Convolution stamps a copy of the response starting at each input sample, so the swell builds over the impulse response's length and peaks at its end — which lands after the dry sound rather than before it.
You get a backwards-sounding texture in roughly the right shape, in the wrong place.
The Fix
Delay the dry signal by the length of the impulse response, so the swell has somewhere to happen before the sound lands, and extend the render by the same amount so the lead-in is not clipped off the front of the file.
That second part matters more than it sounds. A pre-verb whose build has been truncated at the file boundary loses the entire effect, because the build is the effect.
5. Using It Musically
A few things that make the difference between a swell that works and one that sounds like a mistake.
- Length sets the meaning. A short lift of a few hundred milliseconds reads as emphasis on a word or a hit. Several seconds reads as a cinematic build into a section.
- It is an anticipation cue, so put it before things worth anticipating. On every snare it becomes wallpaper; on the one entering the chorus it does real work.
- Isolated sounds are far better subjects than dense mixes. A single vocal line or one drum hit gives the swell a clear target, whereas a full mix produces overlapping swells that just read as wash.
- Everything shifts later. The lead-in delays the dry signal, so a pre-verbed element no longer lines up with material that was not processed. Account for that rather than fighting it.
- It can be combined with a normal tail. Pre-verb into the sound and ordinary reverb out of it is a strong effect for transitions, though it uses a lot of space in a mix.
The Short Version
Reverse reverb is not a special kind of reverb. It is an ordinary reverb whose room recording has been played backwards, and the four-step DAW ritual is a longer path to the same destination.
Once you see that the effect lives in the impulse response rather than in the procedure, the workflow problem disappears: no bouncing, no intermediate files, and a preview that matches the export because there is only ever one render. Hear it on your own audio.
Method
The identity was verified by implementing direct convolution and both routes in JavaScript and comparing outputs sample by sample; the maximum absolute difference of 1.49 × 10⁻⁷ is consistent with 32-bit float rounding rather than any difference in result. The impulse response energy figures come from measuring this site's own `createImpulseResponse` output over a 1.5-second response and summing squared samples in each half.