Impulse Response Room Capture: Record Room Impulse Responses
Overview
Impulse Response Room Capture. Your speakers play a sine sweep, your microphone records it back, and the Impulse Response Room Capture tool cross-correlates the two to derive your room's actual impulse response — the same technique professional acousticians rely on, distilled into a single Play Sweep & Capture button. Run it after the RT60 test if you want a deeper look at how your room behaves across the full frequency range. The vocal range test runs entirely in your browser via the Web Audio API — nothing is uploaded anywhere.
Use impulse response room capture to record how your room sounds, then reuse that sound in any mix. Playing a sweep through a speaker and recording it gives you the raw material for a free, accurate IR, a short audio file that stores every reflection. This guide covers the impulse responses you can make, the sweep settings to choose, and how to turn the raw recording into a file your software can load.
Impulse Response Room Capture Basics
A room impulse response is the recorded answer to a single click. A loud, instant sound is played in the room, and everything the mics hear afterwards, from the direct sound to the early reflections and the fading reverberation, makes up the response. Surfaces bounce energy back while absorption soaks it up, so every space in your studio, hallway or hall leaves its own pattern in the acoustics. Each impulse response you record is tied to one source position and one mic position, so a single room can yield many IRs. The online audio latency check takes under a minute and needs nothing but your browser.
An IR only works when the thing you measure behaves in a linear and time invariant way, usually shortened to LTI. A fixed EQ, a delay, a filter and a static phaser all qualify. A moving flange or chorus is not time-invariant, and compression changes with the level of the input, so one IR cannot hold any of them.
Why Convolution Reverb Needs an IR
A convolution reverb multiplies every sample of your dry signal by every sample of the IR, which lets the room reverberate that signal just as it would in person. Engineers load a captured file into a convolver to match ADR to location sound in post-production, to add ambience to a dry snare, or to place a close-miked part in the studio where it was never recorded. Mixing with a measured space is more believable than a preset, because the decay comes from a real place. Without convolution, a plugin has to fake the tail with delay lines, which is why a measured file sounds more natural. Libraries of impulse responses sell for money, but your own cost only an afternoon, and an impulse response capture of a bucket or a hallway gives your sound design a tone that no preset library has.
Recording Impulse Responses With a Sine Sweep
To create an impulse response from a real room, play a known test signal through a speaker and record what the room does to it. The test signal can be generated in seconds, and a clean test signal is what makes the later math work; each pass yields one impulse response per mic position. A sweep glides from low pitch to high pitch and delivers far more energy than a single click, so the signal-to-noise ratio is much better. This is the standard way of recording impulse responses, and every sine-sweep capture needs one extra step afterwards to compress the response into a usable file. The frequency identification ear training exercise runs entirely in your browser — no account or download needed.
Speaker and Microphone Setup
Pick one speaker and keep the signal chain short, because every extra device colours the result. Put it where the sound source would sit, such as a singer's spot or an amp's corner, and place one microphone or a pair of mics at the listening position. Use the same audio interface for playback and recording so both share one clock source; two devices drift apart and smear the response.
- Condensers are the usual choice for mics, but a fixed height and distance matters more than polar patterns.
- Record in mono or stereo, or in a multichannel or ambisonic format if your loader can load it.
- Leave plenty of headroom, and note the location, angle and mic position for later.
- Test the hardware once at low level before the real take.
Exponential Sine Sweep Method
The exponential sine sweep spends equal time on every octave, which suits how rooms behave. Its convolutional inverse is simply the sweep reversed with a volume ramp, so convolving your recording with that inverse collapses the room's response into a sharp impulse followed by the reverb. A bonus is that the sweep time-separates harmonic distortion from the linear response, so the non-linear artefacts of the playback system land before the IR and can be deleted.
Transient Method With a Balloon
If you cannot carry a speaker, a transient is the direct way to record an IR. A starter pistol, a popped balloon or a clapper board produces a broadband burst, and the recording can be used as an IR with no further processing. Fire the burst at the source position and record it with your mics at the listening position. The trade-off is energy: a burst is brief, so the tail sits closer to the room's own hiss than a sweep would.
Sample Rate and Pre-Generated Sweeps
Match the sample rate of the sweep to your session, because a mismatch shifts the timing and ruins the alignment. Record at 24 bit and save as WAV so every loader can open the result, since WAV is read by nearly all convolvers. Pre-generated sweeps are ready-made files you can play without building your own, which is handy when you want room impulse response capture to take minutes instead of an afternoon.
Capturing a Stairwell: IR Settings in Practice
I wanted a tight, natural decay for a sound project, and the concrete stairwell behind my building was the obvious source. The question before pressing record was simple: how long does the sweep need to be, and how long do I keep going afterwards? Planning the sweep like this is the first step in capturing impulse responses of any venue, and the planner turns three guesses into one recording length.
I measured the space at 58.4 m³ and guessed about 2.9 m² of absorption, mostly the door and some wooden rails. Sabine's formula, 0.161 × V ÷ A, gives 0.161 × 58.4 ÷ 2.9 = 3.24 s. I rounded the expected tail up to 3.5 s. My interface runs at 44.1 kHz, so I entered a low cutoff of 30 hertz, an end of 18 kHz and a sweep length of 12 s.
The planner returned 9.23 octaves at 0.77 octaves per second, a sweep of 529,200 samples, and a minimum measurement of 15.5 s (12 + 3.5), roughly 1.96 MiB as a 24-bit file. That final number is the one I watch, because ending the take at 12 s would have chopped off the very decay I was there to keep.
Then I reran with a shorter sweep to see what the stairwell's traffic would cost me:
| Sweep length | Octaves per second | Minimum measurement | Hiss vs 12 s |
|---|---|---|---|
| 12 s | 0.77 | 15.5 s | baseline |
| 10 s | 0.92 | 13.5 s | slightly worse |
Saving 2 s is not worth a hissier tail in a space where the quiet end of the decay is the whole point, so I kept 12 s. Two cautions came out of this. First, Sabine's formula assumes a diffuse space, and a stairwell is anything but; treat 3.24 s as a planning value and read the real decay from the finished file, trimming any late section that has sunk into the hiss. Second, if the sweep is built at 44.1 kHz but my session runs at 48 kHz, the sweep plays 8.8% fast and the 3.24 s tail would land near 2.98 s. So I confirmed both sessions ran at 44.1 kHz before recording, then read the real decay from the finished file and compared it with my 3.24 s estimate.
Room IR Capture, Deconvolution and Noise
Every successful impulse response room capture comes down to three decisions: the sweep settings, the processing step, and how well you tame the room's hiss. Each room IR also has its own quiet floor, so the subsections below follow that order and finish with a worked example that you can reproduce with any IR capture workflow.
Test Signal Settings and Measurement Length
Three settings define the test signal: the start frequency, the end frequency and the duration. Start below the lowest note you care about, such as 25 Hz, and end near the Nyquist frequency, which is half the sample rate. Any impulse response creator, including the sweep planner tool on this page, lets you generate the file after these three values are set. The recording, called the measurement, must run longer than the sweep by at least the length of the room's tail.
Take a living room captured at 48 kHz, with a sweep starting at that same low note and ending at 20 kHz, lasting 8 seconds, and a reverb tail of about 1.9 seconds:
$$N=\log_2\frac{f_{end}}{f_{start}}=\log_2\frac{20000}{25}\approx 9.64\ \text{octaves}$$
$$R=\frac{N}{T}=\frac{9.64}{8}\approx 1.21\ \text{octaves per second}$$
$$T_{meas}\ge T+L_{IR}=8+1.9=9.9\ \text{s}$$
The sweep holds 384,000 samples at that rate, and the measurement must hold at least 475,200, about 1.36 MiB at 24-bit mono. Raising the sweep time lowers the sweep rate and improves the clarity of the tail, and the table shows the trade for three lengths.
| Duration | Octaves per second | Samples at 48 kHz | Gain vs 6 s (dB) |
|---|---|---|---|
| 6 s | 1.61 | 288,000 | 0 |
| 8 s | 1.21 | 384,000 | +1.2 |
| 12 s | 0.80 | 576,000 | +3.0 |
Longer is not always better: a long sweep gives passing traffic more time to cross your recording, so pick the shortest sweep that reaches your clarity target. The IR length you keep afterwards should cover the full decay, and the frequency response of your monitors still shapes everything below 60 Hz.
Processing Options Explained
The processing software divides the recording by the sweep to reveal the room. Most tools expose a handful of switches:
- Wiener deconvolution resists noise better but can add artefacts when the speaker distorts, so try both modes.
- Auto align repairs a measurement that started early; disable latency compensation in your recording software to avoid creating that problem.
- Normalize scales the peak amplitude to 1, which most loaders expect.
- Fade out softens the very end of the file, and trim removes silence at the edges.
- Smoothing trades accuracy for fewer artefacts when matching tone.
Background Noise and Averaging
Background noise hides the quietest part of the decay. In a small apartment on a busy street, sirens and an elevator leak into the capture and bury the end of the reverb tail. Capture at a quiet hour, and repeat the sweep four times. Averaging four aligned takes cuts random hiss power by about 6 dB, which is the same as doubling the sweep length twice, and it lowers the noise floor. It also lets you throw away a take ruined by a car alarm.
Exporting an IR to a DAW Plugin
Export the finished file as WAV, or AIFF if your loader prefers it, and drop it into the loader of your DAW. Many hosts ship a built-in convolution loader, and third-party options such as Altiverb and Space Designer read the same audio. Name the WAV after the room and mic position so your IRs stay searchable, then load the impulse response. Check by playing a short clap through it: the sound should seem to move into your room.
Creative Uses for Your IRs
Your IRs do not have to describe rooms. Run a signal through a guitar amp and speaker cabinet and the resulting cabinet IR lets you play through that tone at 3 a.m. with no mic setup. Tone matching builds an IR from two similar recordings, which captures only the EQ difference, never the reverb. Converting an IR to minimum phase can reduce latency, and longer files hold more detail in the low frequencies, down to infrasound if you push the start frequency lower. The same sweep recipe applies to a hallway or stairwell captured for sound design, so each new space becomes a reusable IR. Plot the spectrum of any new IR before you trust it, and keep a notebook with each one so you can tell a hallway from a stairwell months later.
Frequently Asked Questions
- What is a room impulse response?
- It is a recording of how a space reacts to a brief, loud sound. The direct sound, early reflections and fading reverb together act as an acoustic fingerprint that a convolution reverb can reuse on any dry audio.
- Should I use a sine sweep or a balloon pop?
- A sweep carries far more energy, so the result has a better signal-to-noise ratio, but it needs deconvolution afterwards. A starter pistol, balloon or clapper board needs no processing and is easier on location, at the cost of a noisier tail.
- How long should the recording be?
- At least as long as the sweep plus the decay of the room. The planner adds your expected room decay to the sweep duration to give the minimum measurement length.
- Why does the sample rate need to match?
- If the sweep file and your session run at different rates, the timing shifts and the deconvolved IR is stretched or compressed. Use one sample rate for the test signal, playback and recording.
- What should the start and end frequency be?
- Start below the lowest frequency you care about, around 20 to 25 Hz for most rooms, and end near half the sample rate, which is the Nyquist frequency. Anything higher cannot be generated.
- Can an impulse response capture a guitar amp or a compressor?
- It captures anything that is linear and time invariant, such as a room, a speaker cabinet or an EQ. Distortion, compression and time-varying effects like chorus cannot be fully captured by one IR.
- How do I reduce background noise in a capture?
- Record at a quiet hour, use a longer sweep, and average several takes. Each doubling of the sweep length or the number of averaged takes lowers random noise by about 3 dB.
- Where can I load the finished IR?
- Most DAWs include a convolution reverb, and plugins such as Altiverb or Space Designer accept WAV files, so export the normalized result as WAV and load it there.