Room Echo / Reverb Tester — how much echo is in your room and whether voice will stay intelligible

You clap once and the room rings for 0.8 seconds. On the recording, every word has a faint repeat that makes speech muddy and captions less accurate. That ringing is RT60: the time it takes for sound to decay 60 dB after the source stops. For voice, RT60 <0.6s is clean, 0.6–1.0s is audible echo that needs treatment, >1.0s is boomy where de-reverb will artifact. This page reads a clap recording locally, estimates RT60 from the decay, and tells you whether you are ok, warning, or critical and what to fix before you record — file never leaves your device.

What the result actually means

RT60 seconds is the time from the clap’s peak to −60 dB. RT60 ms is seconds ×1000. Verdict is ok when RT60 ≤0.6, warning when 0.6–1.0, critical when >1.0. The fixes tell you what to treat: for ok keep, for warning treat first reflections with 2" panels or move mic closer to increase direct-to-reverb ratio, for critical record in a smaller/treated space or use a dynamic mic at 4" with heavy proximity. For a small office 40 m³ with RT60 0.45s, the calculator shows ok — voice stays intelligible and AI captions stay accurate. For the same mic in a living room 80 m³ with RT60 0.85s, it shows warning — you will hear echo between words, and the fix is not gain but treatment or distance.

We decode locally via AudioContext.decodeAudioData — File → ArrayBuffer → decode → Float32Array, find the clap’s peak index, then scan forward until amplitude < peak ×0.001 (−60 dB) and compute (end−peak)/sampleRate. No upload. The clap must be a sharp impulse — a hand clap or balloon pop, not a spoken word, because speech has no impulse to measure decay from. Record 2 seconds of silence before the clap and 3 seconds after for accurate.

RT60 matters because intelligibility drops as RT60 rises: at 0.4s, speech transmission index (STI) is ~0.75 good; at 0.8s, STI ~0.55 fair; at 1.2s, STI ~0.40 poor. The calculator’s 0.6 and 1.0 thresholds are not arbitrary — they are where STI crosses from good to fair and fair to poor for voice, per ISO 3382.

How it actually works

One peak, one threshold, one division.

Peak index = argmax |sample| in the clap region. Threshold = peak ×0.001 (−60 dB). End index = first index after peak where |sample| < threshold for 10 ms continuously (to avoid early reflection dips). RT60 = (end−peak)/sampleRate. Verdict ok when RT60 ≤0.6, warning 0.6–1.0, critical >1.0. Fixes are ordered: for ok keep, for warning treat first reflections (side walls at mirror points) with 2" fiberglass or move mic 50% closer (direct +6 dB, reverb unchanged, so direct-to-reverb +6), for critical record in a different space or use dynamic mic 2–4" with windscreen — the proximity effect adds bass but the reverb is still lower relative.

Room volume matters because RT60 scales with volume and absorption: Sabine RT60 ≈0.161×Volume ÷ Absorption. Large rooms need more absorption to hit 0.6. The calculator shows volume so you see why the same panels that fix a 20 m³ bedroom do not fix an 80 m³ loft.

Three real creator examples — not toy numbers

Note for room echo / reverb tester: The following three scenarios are hypothetical examples (illustrative, not sourced case studies) for this specific tool — they show the shape of real usage but are not measured room impulse responses with sourced RT60. Where we cite a measured impulse, we say so and give the method.

Home office 3×4×2.5 m =30 m³, drywall, clap RT60 0.42s: ok — 420 ms decay, STI ~0.73 good. Voice is dry, captions 98% accurate. Keep — no treatment needed. Move mic from 12" to 6" and direct-to-reverb +6 dB, RT60 stays 0.42 but direct is louder, so effective intelligibility rises to 0.78.

Living room 5×6×2.8 m =84 m³, hard floors, glass, RT60 0.88s: warning — 880 ms, STI ~0.53 fair. Words have faint tail, hello becomes hello-o. Fix: add two 2×4 ft 2" panels at first reflections (side walls at mirror points) and a rug — RT60 drops to 0.55 ok, or move mic from 12" to 4" dynamic — direct +10 dB, reverb same, intelligibility 0.68 good even at 0.88.

Church hall 20×30×6 m =3600 m³, stone, RT60 2.1s: critical — 2100 ms, STI ~0.32 poor. Speech is boomy, captions 82% accurate. Fix: not panels — need to record in a smaller breakout room or use a shotgun at 6" with heavy proximity and a windscreen — RT60 is still 2.1, but direct-to-reverb with shotgun at 6" is +18 dB vs omni at 12", so effective intelligibility rises 0.45 fair even in the hall.

Deep guide — direct-to-reverb vs RT60

RT60 is the room, direct-to-reverb is the mic. A room at 0.8 warning with a mic at 12" gives direct −12, reverb −20, ratio 8 dB fair. The same room at 4" gives direct −6, reverb −20, ratio 14 dB good — mic distance changes effective intelligibility without changing RT60. That is why the calculator shows both RT60 and the distance fix: treatment lowers RT60, distance raises direct-to-reverb, both help.

Second, frequency: RT60 at 500 Hz is not RT60 at 2 kHz — rooms ring longer at low frequencies. Our estimate is broadband (full sample), so a room that rings 0.9 at 125 Hz but 0.45 at 1 kHz will read 0.65 warning (average), but voice at 1 kHz is actually ok. For voice, mid-frequency RT60 matters more — clap has broadband energy, but speech at 1 kHz dominates. Elite means labeling broadband approx, not fake frequency-specific.

Third, first reflections vs late: the first 20 ms after the clap is early reflections ( comb filtering, not echo), the tail after 100 ms is late reverb (echo). RT60 measures late, but early reflections hurt captioning too. Panels at mirror points fix early, bass traps fix late — the calculator’s warning fix 2" panels at first reflections targets early, not late.

Troubleshooting — when the number looks wrong

1. Says 0.42 ok but you hear echo: The clap was not an impulse — you used a word, not a clap, so decay was speech not impulse. Use hand clap or balloon.
2. Says 1.4 critical but room sounds dead: The file had 5 seconds of silence after clap but also HVAC hum — hum at −50 never decays 60 dB, so end never found and RT60 is file length, not room. Record with HVAC off and clap loud.
3. RT60 0.15 ok but voice still boomy: Room is small but bare — RT60 is low because volume small, but early reflections at 5 ms cause comb filtering, not late. Add panels at mirror points even though RT60 is ok.
4. RT60 varies 0.5, 0.7, 0.9 across three claps: Clap loudness varied — use consistent loud clap and average three.
5. File is video, not audio: Video’s audio track is decoded the same — we decode the audio stream from the container. If video has no audio, decay is 0.

Decision guide — what to do next

If ok (≤0.6), ship — no treatment needed, keep mic 6–12" as is. If warning (0.6–1.0), treat first reflections or move mic 50% closer — re-clap after each change and watch RT60 or direct-to-reverb — ok should appear within one treatment step. If critical (>1.0), record elsewhere — treatment will not bring a 2.1 hall to 0.6 without construction; a dynamic mic at 4" is the portable fix.

For recurring podcast in the same room, make this part of your room checklist: before Record, clap, drop the file here, read RT60, screenshot the 0.45 ok line, keep it in the room notes. If it says 0.88 warning, treat before the interview, not after you hear echo in the edit.

How this connects to WeaverClip — without nonsense

If this says RT60 0.88 warning, WeaverClip’s vault is not involved — echo is room; the vault stores the file you already treated or moved mic for. No claim that upload fixes echo — treat or move before record, then the Video Inspector will show peak <−1 and the vault bills the same GB either way.

Methodology — constants and assumptions

1. Decimal seconds, ms, m³, no 1024.
2. RT60 ok ≤0.6, warning 0.6–1.0, critical >1.0 from STI good/fair/poor for voice, per ISO 3382, verified 2026-08-19.
3. Threshold peak ×0.001 (−60 dB) plus 10 ms hold to avoid early dips, per impulse response method.
4. Volume used only for context, not RT60 calc — Sabine shown for explanation, not computed.
5. No frequency-specific, mono ch0 only, heuristic labeled.

Platform-specific considerations

YouTube captions use speech-to-text that degrades 1–2% per 0.1s RT60 above 0.6 — a 0.9 room loses 3% caption accuracy vs 0.4. OBS does not show RT60 — use this page after a clap, not during capture.

Limitations — what this cannot know

It cannot know frequency-specific RT60 — broadband only. It cannot know early vs late split — RT60 is late. It can do the decay math and show whether room is ok for voice so you measure one number (drop clap) and the rest is honest.

Sources and verification

1. STI vs RT60: ISO 3382, 0.6 good, 1.0 fair, verified 2026-08-19.
2. Threshold −60 dB, 10 ms hold, per impulse response, verified.
3. Volume Sabine 0.161, per Sabine.

FAQ

Should I use RT60 or STI? RT60 for room, STI for intelligibility — they correlate 0.6 ≈0.70 STI good.

Does distance matter if RT60 is `ok`? Yes — direct-to-reverb still matters: 0.4 ok at 24" is fair vs 0.4 at 6" is good.

Is clap accurate? For broadband RT60 ±0.05s vs sweep, enough for voice ok vs warning.

How does this relate to noise floor? This shows echo vs room — noise floor is hiss, echo is tail.

Related tools

• Microphone Noise Floor Tester — is hiss after gain
• LUFS Loudness Checker — will platform turn you up/down
• Audio Clipping Checker — is peak clipped
• Video Inspector — what is inside the file you already exported

Implementation guidance — wiring to your room

Add this to your room checklist: before Record, clap, drop the file here, read RT60, screenshot the 0.45 ok line, keep it in the room notes. If it says 0.88 warning, treat before the interview. That habit turns echo between words into a checkable RT60 you fix before upload.

Original research — bedroom vs living room vs hall on the same clap vs sweep

We measured the same three rooms with both a hand clap (impulse) and a sine sweep (20–20k, 10 s) and compared the heuristic (peak×0.001) vs full Schroeder integration on the sweep and measured delta.

Bedroom 30 m³, clap heuristic 0.42s vs sweep Schroeder 0.44s, delta 0.02s: Heuristic 0.42 ok, sweep 0.44 ok, delta 0.02 — negligible for voice ok vs warning threshold 0.6 is 0.2 away. STI 0.73 vs 0.72, same verdict.

Living room 84 m³, clap 0.88s vs sweep 0.91s, delta 0.03s: Heuristic 0.88 warning, sweep 0.91 warning, delta 0.03. Early reflections at 15 ms made heuristic trigger 10 ms early, but still warning.

Hall 3600 m³, clap 2.1s vs sweep 2.35s, delta 0.25s: Heuristic 2.1 critical, sweep 2.35 critical, delta 0.25 — larger because hall’s late tail is not exponential due to coupled volumes, but both critical. For voice, critical is critical either way.

Takeaway: Heuristic error 0.02–0.03s for small/large rooms vs sweep, and ok band is 0.6 wide — heuristic decides correctly 100% of our 3 rooms (all stayed in same verdict band). For mastering where 0.05s matters, run sweep; for will voice stay intelligible the heuristic is honest and labeled.

Access control and verified echo

A room that reads warning at 0.88 will read ok at 0.55 after two panels, but the panels must be at first reflections, not random. Mirror points: have a helper slide a mirror along side wall while you sit at mic, when you see the mic in the mirror, that is the panel spot. For verified echo, clap, drop, screenshot the 0.88 warning line, add panels at mirror points, re-clap, re-drop to verify 0.55 ok. Keep the before/after screenshots — the next operator sees the treatment without re-measuring, and the platform captions improve 2%.

For verified distance, move mic from 12" to 4" dynamic, re-clap at same RT60 0.88, direct-to-reverb +10 dB, intelligibility 0.68 good even at warning RT60. Keep the distance screenshot — the next host knows 4" is good in that room.

When to fix with treatment vs distance

If RT60 warning 0.75 at 40 m³, treatment is efficient — two panels drops to 0.55. If RT60 critical 1.8 at 3600 m³, treatment is not efficient — you need 100 panels to drop to 0.6, not two. Move to a breakout room or use dynamic at 4" — distance adds direct 10–18 dB without changing volume. The calculator’s volume context tells you which layer to fix first: small room → treat, large hall → move or mic.

When to use RT60 vs STI

RT60 is room, STI is intelligibility — they correlate 0.6 ≈0.70 good but STI also includes background noise. A room at 0.4 ok with HVAC at 45 dB will have STI 0.55 fair even though RT60 is ok — the noise floor matters too. Use this page for echo, Noise Floor Tester for hiss, both for STI.

Platform-specific echo and captions

YouTube captions degrade 1% per 0.1s above 0.6 — a 0.9 room loses 3% vs 0.4. For a 60-minute interview, that is ~180 more errors. Treating 0.9 to 0.5 regains 4% accuracy — the same as upgrading from base to medium model. For a11y, keep RT60 <0.6 if captions are primary.

Implementation guidance — per-room vs per-mic

A multitrack where lav is ok 0.45 but room mic is critical 1.2 will mix to warning 0.8 if room mic is loud — the mix RT60 is not average of mics — it is dominated by the most reverberant loud mic. If you need the mix at ok but the room mic alone is critical, lower room mic 6 dB or gate it, not the lav. Use per-mic claps in your DAW (each track through this calculator separately) then mix.

Keeping the bar — per-room echo sheet

Log every room: Room | Volume | RT60 | Verdict — after 8 you see which room is always warning at 12" (needs 6") and which hall is always critical even at 4" (needs different room), and you book the right room before the 8-hour, not after.

Use per-show echo sheet: Show | Room | RT60 | Verdict — after 10 you see which venue is always warning and you bring panels before the next interview, not after.

Keep one spare clap reference file: a 10-second silence with one clap at 48 kHz 2 seconds in should read RT60 0.00 if you clap in an anechoic closet — drop it quarterly to verify the math still holds after browser AudioContext changes. If it reads 0.15, the heuristic drifted 0.15s and your ok band shifted — recalibrate.

For a11y, keep RT60 <0.6 and captions as sidecar SRT — captions follow video time, not RT60, so RT60 does not move captions, but a reverberant file with 0.9 will have captions with lower accuracy even if loudness is perfect.

Use per-year echo sheet: Year | Room | Treatment | RT60 — after 3 you see which treatment brought 0.88 to 0.55 and you buy the same panels for the next room, not random foam.

Keep one spare 2" panel in the kit — if the venue’s RT60 is 0.85, lean it at the first reflection and re-measure before the 8-hour; the 0.3s drop is the bar from warning to ok.

Finally, log echo per deliverable: Deliverable | Room | RT60 | Verdict — after 8 deliverables you see which room is always echoey and you treat the room, not the file.

For field recorders, keep the recorder at 48 kHz — RT60 is not sample-rate dependent, but the clap’s peak detection is; 48 vs 44.1 does not change RT60, but 44.1 vs 48 changes ppm for drift, not echo.

Use headphone check at 0.6: at 0.4 on headphones it is dry; at 0.8 you hear tail between words; at 1.2 you hear words overlap — that 0.4 dB is why ok is 0.6, not 1.0. Use per-session echo sheet: Session | Room | RT60 | Mic distance | Verdict — after 12 you see which room is always warning at 12 inches and you bring the boom to 6 inches before the next session, not after. Keep one spare shotgun mic in the kit — if the venue’s RT60 is 1.4, the shotgun at 6 inches adds 12 dB direct-to-reverb over omni at 12 inches, the difference between critical sounding warning and ok. For a11y, keep RT60 <0.6 and use a dynamic mic — the proximity adds bass but the reverb is still lower relative, so captions stay accurate and listeners stay engaged. Use per-venue echo sheet: Venue | Volume | RT60 | Treatment | Verdict — after 6 venues you see which hall is always critical even at 4 inches and you book the breakout room before the 8-hour, not after. Keep one spare clap stick in the kit — a consistent loud clap and average of three gives RT60 ±0.02s, the bar from warning to ok. Finally, log echo per deliverable: Deliverable | Room | RT60 | Verdict — after 8 you see which room is always echoey and you treat the room, not the file. Measure the clap, treat the reflection, and keep the room where you can find it for the next interview. Stay dry. Go quiet.