Bluetooth microphone latency test — your real round-trip delay

Most Bluetooth speakers come back between 120 and 220 ms on this test, and the figure above is yours rather than a category average. The page plays a 30 ms chirp, records your microphone at the same moment, and cross-correlates the two to find the gap. Five runs, a median, room air time subtracted.

Reading your result

The verdict names a band, and the boundaries sit where the delay changes what you can do:

• Under 40 ms. Effectively no radio in this path. A cable, or the laptop's own speakers.
• 40 to 120 ms. A good AAC link, LE Audio, or aptX Low Latency. Speech feels natural.
• 120 to 220 ms. Ordinary AAC or SBC. You hear yourself a beat behind. Not a fault.
• 220 to 400 ms. SBC on a congested 2.4 GHz band, or a speaker with a deep buffer.
• Over 400 ms. Usually a TV, soundbar, or multi-room group. Unusable live.

Published thresholds put roughly 20 ms below notice, 40 to 50 ms where lips and voice visibly disagree, and past 100 ms in the territory of a bad phone call. Self-monitoring is far more sensitive than passive listening. More on which delays people actually notice and where the milliseconds go between your mouth and the speaker cone.

How the test works

A chirp survives a room in a way a click does not: 30 ms of Hann-windowed sweep from 1 kHz to 6 kHz, a band nearly every portable speaker and phone microphone handles well. Correlating the recording against the chirp that was sent gives a sharp peak at the arrival time, and the height of that peak is a confidence score.

Five runs are combined with a median rather than a mean, and any run below 0.35 confidence is discarded instead of averaged in, because a failed detection is not a fast link. The pills show each run, so a wide spread stays visible. Sound travels about 2.9 ms per metre, so the distance you pick is subtracted, leaving the electronics rather than the room.

What published codec figures say

Sources genuinely disagree. Showing that is more useful than averaging it into one confident-looking number.

iOS exposes no codec picker, so you cannot ask for anything faster, and the low-latency codec everyone recommends is not on the phone at all. If your number sits well outside the AAC range, the extra is the speaker's own buffer, which the codec comparison covers source by source.

What this test cannot tell you

It measures a loop, not a one-way trip: output buffering, the Bluetooth encode and decode, the speaker's buffer, the air, and your microphone's input path, all in one number. Isolating the speaker's share needs a second device of known latency to subtract.

It also measures this browser, not the app. Safari's Web Audio output and an iOS audio session are different code paths with different buffer sizes, so the local end can differ by tens of milliseconds either way. The Bluetooth portion of the number is the same whichever software feeds it, because the codec, the radio hop and the receiver's own buffer sit outside the phone's audio stack entirely. What differs is the buffering on the phone side, in the input and output paths the browser and a native app each set up for themselves, so expect the two totals to land close to each other rather than exactly on top of each other.

And a result is a snapshot. Congestion on 2.4 GHz moves it, so a number taken at a quiet desk is no promise about the same speaker in a full room.

Nothing on the phone removes this

No app can remove the delay you just measured, and any app claiming otherwise is either wrong or quietly means a cable. It is created downstream of every app, in the codec and the receiving speaker's buffer, and iOS grants an app no control over either. The category will not say that: the "no lag" line in these listings is the most repeated false claim in the niche.

What to do with your number

In the 120 to 220 ms band, stop listening to yourself: turn the speaker away from whoever is talking and the delay stops fighting their voice. For announcements it barely matters. For singing it matters a great deal, which is the karaoke problem, where the backing track and the voice arrive at different times. Above 400 ms, check what you are connected to before blaming the radio, and confirm the speaker with the channel, tone and sweep checks. Then hear what your delay feels like on the live monitor. If it is unusable, run a cable: wired is the only thing that genuinely fixes the number.