Room Reverb for WAV Files

What it is for

A dry recording, such as a voice recorded close to the microphone, a synthesized sound or a foley clap, often sounds pasted on. A little room reverb places it in a space. This tool adds a stereo reverb with seven room presets: small room, living room, studio, concert hall, cathedral, cave and open forest. You can override the reverberation time (RT60, the time for the reverb to fall by 60 dB) and the levels of the reverb and the original.

Unlike most "room size" knobs, the reverberation time here is a real number. The engine sets the internal feedback so that the decay reaches −60 dB in exactly that time, and then measures the reverb's impulse response to confirm it. The presets are typical values for such spaces. They do not model the geometry or early reflections of a particular real room.

Worked example

We made a 0.4-second burst of white noise (a snare-like hit, 44.1 kHz mono) with the Noise Generator and processed it twice.

Small room with preset values: "Small room reverb (RT60 0.4 s set, 0.4 s measured) on 0.4 s of audio → 0.9 s stereo WAV, peak −2.4 dBFS." The file grew by a 0.48 s tail. RMS fell from −16.8 to −20.6 dBFS because the energy is now spread over more than twice the length, and the spectral centroid barely moved (11,069 → 10,970 Hz).

Cave with RT60 overridden to 4.5 s: "Cave reverb (RT60 4.5 s set, 4.52 s measured) … → 5.8 s stereo WAV, peak −0.5 dBFS." The centroid dropped to 9,077 Hz. The damping lets high frequencies die away sooner than lows, just as stone and air do.

How to read the result

• Reverberation time (set / measured): the measured value comes from Schroeder backward integration of the reverb's impulse response (the −5 to −25 dB part of the decay, extrapolated to −60 dB). It describes low and middle frequencies, and highs decay faster. Expect it within a few percent of the set value.
• Length: with "keep the reverb tail" the file grows by about 1.2 × RT60 plus the pre-delay.
• Before and after: RMS, peak and spectral centroid of the original and the result. A lower centroid means the reverb sounds darker.
• Gain reduction (warning): long reverbs on sustained sounds can build up above full scale. Then the whole file is turned down to a −1 dBFS peak instead of clipping.
• Parameters: the full preset values used, including damping, width and pre-delay.

Tips and common mistakes

• Less is more. For speech, a small room or studio with the preset levels is usually enough. Hall and cathedral suit music, effects and ambience.
• Set original level 0 to get only the reverb, for example to layer it under the dry track in your editor.
• If you get the gain-reduction warning, lower the reverb level (0.25–0.35 for long RT60s) instead of accepting a quieter file.
• Keep the tail on when the sound ends abruptly. Turning it off cuts the reverb at the original length.
• A mono input becomes a stereo output. The width of the preset controls how far apart left and right sound.
• Apply reverb last, after cutting and loudness changes.

FAQ

What algorithm is used?

The Freeverb structure (eight parallel damped comb filters and four all-pass filters per channel, public domain), with each comb's feedback computed from the reverberation time instead of a fixed room-size scale.

Which files can I upload?

WAV directly, and MP3, FLAC, Ogg, M4A or AIFF through ffmpeg decoding. The limit is 120 seconds at 8–96 kHz. The result is always WAV.

Why does the measured RT60 differ slightly from the setting?

The all-pass filters and the stereo spread add a little diffusion. The measurement is also a line fit to a slightly wavy decay. Differences of 1–3 % are normal.

Is the original file kept?

No. It is processed in memory or in a private temporary folder. The result can be downloaded for about 2 hours.