Audio Latency Test: Measure Round-Trip Delay Automatically

How the microphone loopback test works

This audio latency test works without any tapping. When you press Start, the page asks for your microphone with echo cancellation, noise suppression and automatic gain switched off, because each of those can hide or reshape the test sound. It then schedules five short chirps – quick sweeps from low to high pitch, 20 milliseconds long and 300 milliseconds apart – on the browser’s audio clock, and records what the microphone hears in the same clock’s timeline.

To find the chirps in the recording, the page compares the recording with the chirp it sent at every possible time shift (a cross-correlation, computed with a fast Fourier transform). The shift where they match best is the round-trip delay. A sweep is used instead of a click because its match peak is sharp and stands out clearly from room noise. The page then checks that the five chirps came back with the same spacing they were sent with; if they did not, the run is thrown away instead of reported.

That whole cycle repeats five times. You get the median, every individual run and the spread between the highest and lowest run. The arithmetic itself is precise to a single audio sample (about 0.02 ms at 48 kHz); the spread you see comes from the system’s buffers changing slightly between runs, and the result card shows it so you can judge how repeatable your setup is.

Results are estimates of the whole playback chain plus the input chain. They are not a laboratory measurement, and different browsers will report different numbers for the same hardware.

Round-trip vs output latency

The number this test reports is the round trip: output buffers, the speaker or headphones, the sound’s trip through the air, the microphone and the input buffers, all added together. Sound travels about 34 cm per millisecond, so a speaker 30 cm from the microphone adds roughly 1 ms – small, but it is in there.

Output latency is only the first half: how long a sound takes to come out after an app plays it. That is what you notice when video and sound drift apart, and it is what the flash-and-beep audio delay test estimates. A microphone test cannot split the round trip into its two halves on its own, because it only sees the total. If you need the output part only, use the slider test or the Bluetooth comparison; if you record or perform through your computer, the round trip is the figure that matters, because you hear yourself after both halves.

What your browser reports

Next to the result, the page shows three values the browser itself provides. They are reported, not measured, and some browsers leave them out – in that case you will see “not reported by this browser”.

• Sample rate – how many audio samples per second the browser is using, usually 44,100 or 48,000.
• Base latency – the delay the browser’s own audio engine adds before handing sound to the system.
• Output latency – the browser’s estimate of the delay from its audio engine to the speaker. Wireless devices often make this estimate less reliable.

The measured round trip is usually larger than these values combined, because they leave out the input side, and because the system and drivers rarely report every buffer they use.

Tips for an accurate reading

• Use the built-in microphone of your laptop, phone or tablet, or a desk microphone; do not use the microphone on a headset.
• Play the chirps through speakers. If you want to include headphones, hold an ear cup or earbud right next to the microphone.
• Set the volume to about half; too loud can distort, too quiet can drown in room noise.
• Keep the room quiet and the device still for the whole measurement, about 15 seconds.
• Close other apps that use audio or video calls, so nothing else has hold of the microphone.
• If the page warns that audio processing is still on, try another browser: the warning means the reading may be too high or fail.

Limitations

This audio latency test cannot separate input delay from output delay; it always reports the sum. It also cannot see inside your headphones: a headset microphone usually switches wireless headphones into a call mode with different sound quality and timing, so it would measure that mode rather than normal listening. Some phones keep echo cancellation on even when the page asks for it to be switched off; the page warns you when it can tell, but it cannot always tell. Finally, the test measures your browser’s audio path. A music app with its own driver may be faster or slower on the same hardware.

How to reduce audio delay

Start with the cheapest fix: the one that removes a step from the chain. Measure again after each change so you know which one helped. Menu names below vary by brand and system version, so look for similar words if yours differ.

Windows

• Turn off sound enhancements and spatial sound for your output device in the sound settings.
• Use the device’s own driver rather than a generic one; for music software, an audio interface with a dedicated low-latency driver usually helps most.
• Close other apps that are playing or recording sound.

macOS

• Use wired headphones or the built-in speakers when timing matters.
• Check the sound output in System Settings and avoid routing audio through extra virtual devices.
• In music software, lower the buffer size step by step until you hear crackles, then go one step back up.

Android

• Turn off sound effects or equaliser features you do not need.
• If your headphones have a companion app, look for a game or low-latency mode; the name varies by brand.

iPhone and iPad

• Turn off spatial audio or head tracking for the headphones if you do not need it.
• Use the headphones’ app, if it has one, to check for a low-latency or game setting.
• For games and music apps, wired headphones through an adapter avoid wireless delay entirely.

TV, soundbar and receiver

• Switch the TV to game mode, which usually cuts picture processing and therefore the picture delay.
• Look for an AV sync, audio delay or lip sync setting and use the number you measured here.
• Set the correction in one device only, ideally the last one in the chain, so two corrections do not fight each other.

Your privacy

This test asks for your microphone only after you press Start. The signal is analysed in your browser’s memory, a couple of seconds at a time, and then thrown away. Nothing is recorded to a file, nothing is uploaded, and the microphone is switched off as soon as the five runs finish. We only count anonymous events such as “a test finished with a result between 100 and 200 ms”, never the sound itself or the exact number.

Details are in the privacy policy.