Record in the same state as a world-class engineer. Anyone, anywhere.
KUON FIELD compares, in real time, what your microphone captures right here against an ideal or a master recording, and reports the match as a number. At the same time it predicts the room’s sound field from physics and shows it in three dimensions. It turns recording acoustics — long locked inside ears and decades of experience — into something measurable, visible and reproducible. We believe this is a first of its kind.
The “Match (live)” mode inside the KUON STAGE hub. No install, no audio upload, runs on your device. ¥9,800 one-time, or included in Concerto (¥1,980/mo).
This is both a user manual and a technical specification. The first half is the clearest possible guide to using it; the second half is a technical specification for researchers, with equations and physical standards. Read from wherever you need.
Intent — democratizing recording acoustics.
Great recording has lived inside the ears of a few engineers and decades of experience. Why one position is better than another was rarely put into words — an unrepeatable, unverifiable craft.
KUON FIELD turns that quality into a match — how close you are to a reference — shown live. It also predicts, from physics, how sound spreads through the room. The result: even someone holding a mic for the first time has the same decision basis as a world-class engineer.
We do not grade people with numbers. The match is a distance from a reference; a healthy range is 70–85%. This mirrors our stance elsewhere: a master’s choice is one valid choice, and so is yours. The tool measures; you decide.
What becomes possible
Finding the sweet spot for a soloist
Move the mic in front of a piano or cello and the match traces a peak. The highest point is a good spot for that instrument and room.
Reproducing a master recording
Load a recording you admire and bring your miking toward its tone and space. The audio is reduced to features and discarded.
Placement for ensembles and chamber music
Set the direct-to-reverberant balance and stereo image to a target while checking each as a number.
Total venue acoustic coordination (with KUON PIANO)
Tune the instrument itself optimally with KUON PIANO, then place mics and source while predicting the venue’s field with KUON FIELD. Instrument, miking and field, coordinated as one system — a first-of-its-kind end-to-end flow.
Research and education
Reproducible, quantitative acoustic measurement and prediction, in a browser, without dedicated hardware — for teaching, papers, and quick hypothesis prototyping.
Complete manual — everything, step by step
Even your first session, start to finish without guessing. Each control is explained.
0. Where it lives
Open the app and choose the “Match (live)” tab at the top. That is KUON FIELD. The “Design” tab beside it is the same hub’s design simulator (prediction only).
1. Choose a reference (the target)
Choose what to compare against. Four ways: (a) Set this position — make the mic’s current spot the target; ideal for A/B comparing positions. (b) Built-in profile — a provisional balance baseline (neutral / slightly warm / slightly clear). (c) A saved AUDIO MIRROR reference. (d) Reproduce a master — load a target audio file. All compare tonal shape, not loudness.
2. Start the mic
Press “Start mic” and allow access. For measurement accuracy the signal is captured raw, with echo cancellation, noise suppression and auto-gain all off. A stereo (two-channel) input is recommended.
3. Read the match and the breakdown
Two numbers appear. The instant % is a fast needle that moves with each note (a short window); the integrated % settles over the whole take. Below them, four axes — tone balance, spatiality, stereo width, attack — show which is off and translate into action: move back, widen the spacing.
4. See “when and which band” in the heatmap
Vertical is frequency, horizontal is time, color is deviation from the reference (cool = below / cream = matched / warm = above). The line below is the score history, with the 70–85% healthy band shaded. Save the view of a good position with “Save as image”.
5. Use the predicted field (2D / 3D)
A top-down 2D map and a 3D scene visualize the field. Two layers: “Direct/Reverberant (D/R)” shows close vs roomy, and “HF loss” shows how much the air takes from the highs. Drag the source and mic; change dimensions, wall absorption, temperature and humidity with sliders and the map responds at once. Read the numbers at the mic point (D/R, distance, RT60, critical distance, 8 kHz air loss).
6. Clap to measure the room
Press “Measure the room (clap once)” and, in a quiet space near the mic, clap once clearly. It measures the reverberation (RT60) from the decay and calibrates the predicted map with the real room rather than a guessed wall material. “Clear” returns to the material preset.
7. A recommended workflow
Choose a reference → start the mic → clap once to calibrate → enter temperature and humidity → play while watching the match and breakdown, moving the mic → confirm the peak on the heatmap → save the best view as an image → survey the whole in 3D.
Technical specification — for researchers
Everything below rests on established acoustics and signal-processing standards, and is numerically verified against synthetic signals and published values. There are no hidden approximations or black boxes.
A. Signal processing of the miking match
From the stereo input we continuously extract this “recording fingerprint” (48 kHz internally):
· LTAS (long-term average spectrum) = 1/3-octave band levels in dB (tonal balance) · IACC (inter-aural cross-correlation) = L/R correlation and image width (0..1) · Mid/Side energy ratio in dB = center vs spread · Attack = crest factor 20·log10(peak/RMS) (sharp up close, soft at distance)
We evaluate on two timescales: instant (a short window of about 0.34 s, close to a single attack) and integrated (a running average of windows over the whole take).
levels_norm(b) = levels(b) − mean_b(levels)d_norm = min(1, |Δ| / scale)
score = 100 · exp(−2 · d_norm)
overall = 0.38·tone + 0.27·spatial + 0.17·width + 0.18·attackThe mean absolute dB difference across normalized bands (MAD), and the IACC and Mid/Side differences, become a normalized distance d, then a score. The overall is a weighted blend (tone 0.38 / spatiality 0.27 / width 0.17 / attack 0.18). 100% is not the goal; a healthy range is 70–85% (commercial masters shift LTAS by mastering).
B. Physics of the field prediction (the predicted map)
The room-wide field is predicted with established room-acoustics equations. For each receiver point we compute:
RT60 = 0.161 · V / (S · α)r_c = 0.057 · √(V / RT60)D/R(d) = 20 · log10(r_c / d) [dB]Atmospheric absorption α(f,T,RH): strictly per ISO 9613-1 (oxygen and nitrogen relaxation frequencies). At 20°C/70%RH it matches published values (1 kHz ≈ 5 dB/km, 8 kHz ≈ 78 dB/km). Highs fall off strongly with distance, temperature and humidity.
α(f) = 8.686 · f² · [ classical(T,p)
+ relax_O2(f,T,h) + relax_N2(f,T,h) ] [dB/m]Early reflections use the image-source method; diffuse-field spatial coherence uses sinc(kd); the speed of sound is temperature-dependent, c = 331.3·√(T/273.15).
C. Room measurement (clap → RT60 → calibration)
From the clap (an impulse-like signal) we form the energy decay curve (EDC) by Schroeder backward integration, take the slope from −5 dB to −25 dB (T20), and multiply by three to obtain RT60. On synthetic decays it recovers a known RT60 to three significant figures. We back-solve Sabine for an equivalent absorption and feed it into the map in place of a guessed wall material.
EDC(t) = ∫_t^∞ h(τ)² dτ
RT60 = 3 · ( t(−25 dB) − t(−5 dB) )
α_eq = 0.161 · V / (S · RT60_measured)D. Architecture and the honesty boundary
Everything runs on your device. Mic analysis, reference extraction, recording, 3D rendering — audio is never sent to a server. A loaded master is reduced to features and discarded. The foundation is the browser’s standard capabilities (high-precision audio analysis, a WebAssembly DSP core, WebGL 3D).
We draw the line honestly. The room-wide field is a prediction (one mic cannot sense the whole room). The real measurement is only at the mic’s current point, overlaid on the predicted map and clearly distinguished. Temperature and humidity are entered by hand, as browsers have no such sensor. We never present a prediction as a measurement.
KUON PIANO × KUON FIELD — the venue as one system
KUON PIANO tunes the instrument itself, with a stretch curve derived from that very piano’s inharmonicity (how its strings drift slightly from whole-number partials). It makes the source right.
KUON FIELD then decides how that well-tuned instrument is captured and how it spreads through the venue, via the match and the predicted field. Source → tuning → miking → field, coordinated as one system.
A tuner, a recording engineer and a performer can then share, on one screen, “in this venue, to this seat, deliver it like this.” No dedicated hardware, no server.
All of it, inside your browser.
Mic analysis, reference extraction, recording, 3D — all on your device; audio is never sent to a server. No install. Your sound never leaves your hands.
Frequently asked
Is it really a “first”?
Reference comparison and acoustic simulation exist separately. As far as we know, no other tool offers real-time miking match and field prediction — temperature and humidity included — together, in a browser. We say so within what we could verify, without exaggeration.
What is the “target” of the match?
An ideal profile for the instrument, a built-in provisional baseline, or a master you load. It is not about 100% — a healthy range is 70–85%. The number is a distance from the reference, not a grade.
Is the 3D field measured?
The room-wide field is a physics prediction. The real measurement is only at the mic’s point, overlaid and distinguished. See Technical spec D.
Can temperature and humidity be automatic?
Browsers have no such sensor, so they are entered by hand. Room reverberation, however, can be measured by clapping and feeds the map.
Install or cost?
No install, no audio upload. Buy it outright (¥9,800, updates included) or use it with the Concerto plan (¥1,980/mo). Customers outside Japan see local-currency pricing at checkout.
What mic do I need?
A stereo (two-channel) input is recommended. An omnidirectional AB pair such as our P-86S / X-86S is ideal, but a built-in stereo on a laptop, tablet or phone is enough to start.