Why is there a delay in your audio, and how do you reduce latency on a Mac?

Where every millisecond of audio delay comes from on a Mac: buffer size, Bluetooth, processing and the network. How much latency is normal, a simple clap test to measure yours, and the fixes that work.


You say "can everyone hear me?", wait, hear nothing, and start the sentence again. At that exact moment your colleague says "yes, go ahead", and now you are both talking over each other. Or you put on AirPods to hear yourself while recording in OBS, and your own voice arrives in your ears a beat behind your mouth, so you start tripping over words you have said a thousand times.

Both are audio latency: the time between a sound happening and a sound arriving. It is never zero, and it does not come from one place. It is a chain of small waits, and the trick to reducing it is finding out which link in your chain is the long one. Usually only one of them is.

If your problem is specifically hearing yourself in headphones while you talk, and you want the setup rather than the theory, How do you hear yourself in your headphones while you talk? is the more direct page. This one explains where every millisecond comes from and what you can actually do about each.

Where the delay in your audio comes from

Follow your voice from your mouth to the other person's ear on a normal call from a Mac. Every stage holds the sound for a moment:

StageWhat waitsTypical sizeCan you change it?
Air between mouth and micSound travelingAbout 1 ms per 34 cmMove closer
Converter in the mic or interfaceAnalog to digital conversionWell under 1 msNo
USB transport and driver bufferA block of samples collected before handing overFrom about 1 ms to tens of ms, set by buffer sizeYes, in recording apps
Bluetooth linkCodec plus radio bufferingNot shown anywhere on the MacOnly by switching to a wire
Processing (noise removal, effects, virtual mics)Each processor works on blocks of audioFrom almost nothing to tens of msTurn it off, or choose a lighter tool
Network and the call app's jitter bufferPackets traveling plus a cushion for late packetsUsually the largest share on an internet callWired internet, fewer uploads running
Playback buffer on the other sideSame as yours, in reverseA few ms to tens of msNot by you

The converter figure comes from Focusrite's explainer on drivers and latency, which also names the buffer as usually the biggest factor inside your computer, and often the only one you control. On an internet call, though, the network and the call app's own buffering tend to outweigh everything on your desk.

That distinction decides what kind of delay you have. There are really two problems hiding under one word:

  • Monitoring latency: you hear yourself late. This is a local problem. Every millisecond is on your Mac and your gear, and you can fix nearly all of it.
  • Call latency: the other person hears you late. Most of it is network, and your Mac contributes a minority share.

How much audio latency is normal

Two numbers are useful, and they apply to different jobs.

For conversation, the telecom standard ITU-T G.114 says that if mouth-to-ear delay stays below 150 ms, most applications, speech included, feel essentially transparent. It treats 400 ms one way as the upper limit for general network planning. Above that, people start the "no, you go" dance on every turn. (ITU-T G.114)

For hearing yourself, the tolerance is far tighter, because your skull carries your own voice to your ears with no delay at all, and your brain compares the two. Focusrite's rule of thumb is handy here: sound travels about a foot per millisecond, so 10 ms of monitoring latency puts your own voice about ten feet, or three meters, away from you. The arithmetic is worth doing once, because it explains why the same delay is harmless in one place and maddening in another.

DelayLike hearing yourself fromPicture it as
3 msabout 1 m awaySomeone at arm's length
10 msabout 3.4 m awaySomeone across a living room
30 msabout 10 m awaySomeone at the back of a classroom
150 ms(calls)ITU line for conversation that feels transparent
400 ms(calls)ITU upper limit for network planning

The distance column is just 343 meters per second turned into milliseconds. Nobody minds a friend talking from ten meters away; hearing your own mouth from ten meters away is what makes people lose their place mid-sentence.

So "how much is normal" depends on the job. Thirty milliseconds on a Zoom call is invisible. Thirty milliseconds in your own headphones while you sing can be the whole problem.

The buffer, in numbers

Audio moves through your Mac in blocks. The app waits until it has collected a buffer of, say, 256 samples, then processes and hands them over. At 48 kHz, each block costs time like this:

Buffer size   One block at 48 kHz   Round trip (in + out, before overheads)
32 samples    0.7 ms                about 1.3 ms
64 samples    1.3 ms                about 2.7 ms
128 samples   2.7 ms                about 5.3 ms
256 samples   5.3 ms                about 10.7 ms
512 samples   10.7 ms               about 21.3 ms
1024 samples  21.3 ms               about 42.7 ms

The formula is simple: samples divided by sample rate. Focusrite's explainer runs the same sum and gets a theoretical 1.45 ms round trip for a 32-sample buffer at 44.1 kHz. Real round trips come out longer, because drivers and converters add buffers of their own; MOTU's figure further down shows the gap.

Where do you set it? Not in System Settings. The buffer belongs to the recording app. In Logic Pro it lives under Logic Pro > Settings > Audio > Devices, in the I/O Buffer Size menu, and the same pane shows a Resulting Latency readout for round trip and output. Apple's own guidance in the Logic Pro manual is the right rule for any app: when recording, use the lowest buffer that does not cause clicks, pops or crackles; when mixing, go large, because small buffers put more strain on the processor.

The sample rate and format of the device itself are set in Audio MIDI Setup. If you want to know which of its settings matter for a mic and which are noise, What is Audio MIDI Setup on a Mac, and which mic settings in it matter? walks through it.

Another Apple page, on configuring a connected audio device, adds a rule that can save an evening of troubleshooting: avoid using different devices for input and output. A USB mic for input plus the Mac's headphone jack for output means two separate clocks that have to be kept in step.

Bluetooth audio delay on a Mac

This is the delay many Mac users hit first, and it is the one with no slider.

When you only listen, Bluetooth headphones use the A2DP profile: stereo, decent quality. The moment any app opens their microphone, the headset has to switch to the Hands-Free Profile, which carries mono audio both ways at a low sample rate, 8 kHz or 16 kHz depending on the headset (Microsoft's Bluetooth audio documentation describes the profiles; they are the same Bluetooth profiles on a Mac). Apple says it plainly in its support note: when an app uses the headphones' mic, audio quality and volume are reduced.

For latency, the issue is that the Bluetooth link encodes, transmits and buffers audio inside the headset and the radio stack, and macOS shows you no number for that delay and gives you no setting to shrink it. That has three practical consequences:

  1. On calls, Bluetooth delay adds to the network delay. On a good connection you may never notice it; on a borderline one it can be the part that pushes you into talking over each other.
  2. For monitoring, Bluetooth headphones are a poor choice. Hearing yourself through AirPods while recording in OBS or GarageBand is likely to feel late, because no buffer setting reaches inside the Bluetooth link.
  3. For video, a Bluetooth mic recorded alongside a camera can land out of lip sync, which you fix in the editor or with an offset (see the OBS section below).

The fix is boring and reliable: for anything where you hear yourself, use wired headphones. For the mic side, Which Bluetooth headset has a mic good enough for calls? covers which headsets hold up and why the profile switch matters more than the brand.

USB microphone latency and low latency audio interfaces

USB microphones and interfaces sit on a short, predictable path, which is why they are the usual answer to "how do I reduce audio latency" for anyone who monitors themselves.

The single most useful feature is a headphone jack on the mic or interface with direct monitoring. The mic sends your voice straight to that jack before it ever reaches the Mac, so no computer buffer applies. Logitech describes the Yeti's headphone output as "latency-free" (Yeti product page), and the Focusrite Scarlett Solo (4th Gen) has a dedicated Direct Monitor button. If your USB mic has a headphone jack with direct monitoring, plug your headphones into it, not into the Mac, and most monitoring latency is gone.

Two catches:

  • Turn off software monitoring when you use direct monitoring. If OBS, GarageBand or Logic is also sending your voice back to the same headphones, you hear yourself twice: once instantly, once late. Focusrite's setup guide for the 2i2 gives exactly this warning. In OBS, set the mic to Monitor Off under Edit > Advanced Audio Properties.
  • Direct monitoring is the dry signal. You hear the mic before any noise removal or effects, because those happen on the Mac. That is usually what you want while recording.

What about "low latency audio interface" as a buying criterion? Interfaces do differ, mostly in how good their drivers are at small buffers. MOTU quotes 2.5 ms round trip for the M2 at 96 kHz with a 32-sample buffer (MOTU M2), and that is a useful benchmark: a well-behaved interface can run tiny buffers without crackles. But even Focusrite points out that only an analog monitoring path is truly zero. If you are deciding between two interfaces purely on latency, the one with a hardware direct monitor knob matters more than a millisecond on a spec sheet. For a voice on calls, any decent class-compliant interface stays well below what the call itself adds.

Processing: noise removal, effects and virtual microphones

Anything that changes your sound needs to look at it first. A noise remover cannot decide whether a sound is a keyboard click until it has heard a slice of it; a compressor or equalizer processes audio in blocks. Every processor in the chain adds its own wait, and they stack.

This is where a lot of mystery delay comes from on Mac setups that grew over time: a USB mic, a routing app, OBS filters on top, then Zoom's own noise suppression, then sometimes macOS Voice Isolation, one of the Mic Modes, as well. Each layer may be small, but four of them add up, and two noise removers in a row each take their own cut at the voice. The fix is to keep one processing layer and switch off the rest.

TunedMic is one of those layers, so its cost belongs in this article too. It is a Mac app that adds a virtual microphone tuned to your voice after a one-minute Sound Check, and apps like Zoom, Telegram, Discord or OBS pick it like any other mic. It adds less than a tenth of a second of delay. On a call, that disappears into the connection's own delay. Where it does not disappear is live monitoring of your singing or speech through headphones: for that, keep using your mic's direct monitoring jack, and let TunedMic handle what the other side hears.

Audio latency test: how to measure yours

Guessing wastes time. Three tests, from easiest to most precise:

1. Read the number your app gives you. Logic Pro's Resulting Latency display shows round trip and output latency for the buffer you chose. If your app shows a latency figure next to its buffer setting, read that instead. Either way, this is your local monitoring latency, without Bluetooth.

2. The clap test for monitoring. Put your headphones on, start monitoring in your recording app, and hold your phone's voice recorder right against the headphone cup, with the volume up. Clap once, near the mic. The phone records the real clap and, a moment later, the clap coming back through the headphones. Open the recording in any editor with a waveform, zoom in and measure the gap between the two spikes. That gap is your monitoring latency, give or take a millisecond of air. Do it once with wired headphones and once with Bluetooth ones, and you will know which link is the long one.

3. The clap test for calls. Join the same call from your Mac and from a second device in the same room, such as a tablet or another laptop. Mute the second device's mic so nothing loops back, keep its speaker up, and start your phone's voice recorder. Clap once near the Mac's mic. The recording catches the clap twice: once through the air, once from the second device's speaker. That gap is the full mouth-to-ear delay. Compare it against the 150 ms line from G.114. Repeat on Wi-Fi and on Ethernet; the difference can be bigger than anything you will win inside the Mac.

Fixing audio and video out of sync in OBS

Sometimes latency shows up as lips moving before the voice, or after it. Cameras, capture cards and USB mics each have their own delay, and they rarely match.

OBS has a direct fix: Edit > Advanced Audio Properties, then the Sync Offset column next to your mic, a value in milliseconds. If the voice arrives early relative to the video, delay the audio by the gap you measured (a clap on camera makes it easy to see in a recording). It does not reduce latency, it aligns two delays so they look like none. While you are in that window, check Audio Monitoring: Monitor Off for the mic if you use your mic's headphone jack.

How to reduce audio latency: a checklist by symptom

SymptomMost likely causeWhat to do
You hear yourself late in headphonesSoftware monitoring plus a large buffer, or Bluetooth headphonesWired headphones into the mic's or interface's jack, direct monitoring on, software monitoring off
You hear yourself twiceDirect and software monitoring both onTurn off monitoring in the app
Crackles appear when you lower the bufferThe buffer is now too small for the CPU loadGo one step up, close heavy apps, avoid mixed input and output devices
Callers talk over youNetwork delay plus app bufferingEthernet or a closer Wi-Fi spot, close uploads, check the call app's connection stats
Calls feel slower only with AirPodsBluetooth link plus profile switchWired headset or USB mic for calls
Lips and voice out of sync in a recordingCamera and mic delays differOBS Sync Offset in ms, or shift the audio in your editor
Delay grew after installing toolsStacked processing layersKeep one noise remover, turn off the others

When TunedMic is the right fix, and when it is not

Be clear about what you are solving. If your complaint is delay, TunedMic is not the fix: it adds a little, under a tenth of a second, and nothing in it makes a Bluetooth link or a slow network faster. The cures for latency are the ones above: a wired path, a sensible buffer, direct monitoring and one processing layer instead of four.

Where it fits is the case most people are in once the delay is sorted: the call works, but the voice sounds thin, boomy, distant or noisy. TunedMic measures your voice, mic and room during Sound Check, then serves a tuned version as a separate microphone, so you can switch off the stack of Zoom suppression, Voice Isolation and OBS filters and keep one layer. It runs locally on the Mac. It needs a Mac with Apple silicon (M1 or newer) and macOS 14 or later; there is no Windows or Intel version. There is a free 7-day trial that starts at first launch, with every feature and no card, and no refunds after purchase, so the trial is the time to judge it.

What Sound Check looks like is in the 38-second video on the home page: a minute of reading aloud, then a tuned mic that works in Zoom, Google Meet and Discord. Watch it, then let the trial decide before you check pricing.

What to do right now

Run the clap test before changing anything. If the two spikes are clearly apart and you are wearing Bluetooth headphones, swap to a wired pair plugged into your mic's or interface's headphone jack and turn off monitoring in your app. Then clap again. If the gap is gone, you have found the long link in your chain; if it is still there, lower the buffer one step at a time in your recording app until the clicks start, and go back up one.

Keep reading