Recording Latency: Why There's a Delay and How to Fix It

What Latency Actually Is (and Why It Exists)

Latency is the delay between making a sound and hearing it come back through your headphones or speakers. You sing a note, and a fraction of a second later you hear yourself. If that fraction is small enough, your brain ignores it. If it isn't, playing in time becomes nearly impossible — your timing drifts, your phrasing stiffens, and everything you record feels slightly wrong.

The delay exists because digital audio can't be processed one sample at a time. Here's the round trip your signal makes:

1. Your voice or guitar hits the microphone or input jack as an analog signal.
2. The interface's analog-to-digital converter (ADC) turns it into numbers.
3. The computer collects those numbers into a buffer — a small batch of samples — before processing them, because handling audio in batches is far more efficient than handling each sample individually.
4. Your recording software (the DAW, or digital audio workstation) processes the buffer, running it through any plugins on the track.
5. The processed audio goes into an output buffer, back through the digital-to-analog converter, and finally reaches your ears.

Every stage adds delay. The buffers are the biggest contributors and the only ones you directly control, which is why almost every latency fix starts with buffer size.

How much delay is too much? There's no single magic number — sensitivity varies by person and instrument — but most people play comfortably when the total round trip is under about 10 milliseconds, start to notice somewhere in the 10–20 ms range, and struggle beyond that. Drummers and percussive players tend to notice earliest. For reference, standing about three meters from your guitar amp puts roughly 9 ms of acoustic delay between you and the speaker, and guitarists have coped with that forever. Context matters.

Buffer Size Explained

Buffer size is set in your DAW's audio preferences or your interface's control panel, measured in samples. Common values run from 32 up to 1024 or 2048.

The math is simple: divide the buffer size by the sample rate to get the time one buffer represents. A 256-sample buffer at 48 kHz is 256 ÷ 48,000 = about 5.3 milliseconds.

But that's one buffer, one direction. Real round-trip latency includes the input buffer, the output buffer, converter delay, and driver overhead — so the number you actually experience is typically two to three times the single-buffer figure. Many DAWs display a reported round-trip value in their audio settings; treat it as an estimate, since drivers don't always report honestly.

The trade-off is CPU load. A small buffer means the computer must finish processing each batch in less time. Miss the deadline and you get clicks, pops, and dropouts — the crackling sound of an overworked audio engine. A large buffer gives the CPU breathing room but adds delay.

This leads to the single most useful habit in home recording: use a small buffer while tracking, and a large buffer while mixing. While recording, you need low latency and are usually running few plugins. While mixing, you're not performing, so latency is irrelevant — crank the buffer to 1024 and let your plugins breathe. Changing it takes ten seconds in your audio preferences. Make it a reflex.

The Fix, In Order

Work through these in sequence. Most people solve their problem at step one or two.

1. Use direct monitoring if your interface has one

Most audio interfaces — including nearly all entry-level 2-input models — offer direct monitoring: a hardware path that routes your input straight to your headphones inside the interface, before the signal ever reaches the computer. The delay is effectively zero because there's no buffer involved.

It's usually a knob or button labeled "Direct," "Mix," or "Monitor," blending the direct input signal against playback from the computer. Larger interfaces do the same job through a software mixer utility.

If you're monitoring this way, turn off input monitoring on the track in your DAW. Otherwise you hear yourself twice — once instantly through hardware, once late through software — which produces a doubled, robotic, phasey sound. That doubled sound is one of the most common "latency problems" reported by beginners, and it isn't a latency problem at all; it's two monitoring paths running at once.

The limitation: direct monitoring gives you the raw, unprocessed input. For vocals, bass, acoustic guitar, and DI'd electric that you'll re-amp later, that's fine — plenty of professionals track with a completely dry cue feed. If you need to hear processing while you play, see step 5.

Direct monitoring is genuinely one of the features worth caring about when choosing your first audio interface — far more than converter marketing claims.

2. Lower the buffer for tracking

If you must monitor through the DAW, drop the buffer to 128 samples, or 64 if your system stays clean. Record a test pass and listen for crackles. If you get them, either raise the buffer one notch or reduce the processing load (step 4).

3. Use the right driver (Windows users especially)

On a Mac, the built-in Core Audio system is efficient and this step rarely applies. On Windows, driver choice is often the entire problem.

Windows offers several audio driver paths, and the default ones (often labeled MME, DirectSound, or shared-mode WASAPI) are built for playing videos, not recording — they add large hidden buffers on top of whatever your DAW shows. If your DAW says 128 samples but the delay feels like a quarter second, this is almost certainly why.

The fix is ASIO — a driver standard built for recording that bypasses the Windows audio stack. Install the ASIO driver from your interface manufacturer's website (not the generic one Windows auto-installs), then select it in your DAW's audio settings. Generic ASIO wrappers exist for hardware without a native driver and can work in a pinch, but a manufacturer-written driver is nearly always more stable. This one change routinely takes Windows systems from "unusable" to "fine."

Also check your interface's control panel for a "safe mode" or extra-buffering option, which trades latency for stability. Turn it off while tracking if your system can handle it.

4. Reduce the processing load

Every plugin running in your session eats into the CPU time available per buffer, and some plugins add latency of their own on top — lookahead limiters, linear-phase EQs, and pitch correction are the usual suspects. Your DAW quietly delays everything else to keep tracks aligned (this is called plugin delay compensation), and that added delay applies to your live input too.

While tracking:

• Bypass or remove high-latency plugins from the master bus. A lookahead limiter on the master while you're cutting vocals is a classic self-inflicted wound.
• Freeze or bounce CPU-heavy virtual instrument tracks — most DAWs have a freeze function that renders a track to audio temporarily.
• Many DAWs offer a "low latency mode" or "constrain delay compensation" switch that automatically bypasses high-latency plugins while recording. Learn where yours is.
• Keep the plugins on the record-armed track light: a simple EQ and compressor are fine, a mastering chain is not.

5. The amp sim exception

Recording electric guitar through amp simulation plugins is the one common case where direct monitoring can't help — the raw DI signal sounds nothing like the amped tone, so you must monitor through the computer. This is the scenario that genuinely demands a small buffer (64 or 32 samples), a proper driver, and a light session. If your computer can't run an amp sim cleanly at 128 samples or below, practical workarounds include a hardware modeler or pedal in front of the interface for monitoring, or simply tracking with a slightly compromised tone and re-amping the clean DI afterward.

6. Fix the computer, not just the settings

If clean low-buffer performance still eludes you:

• Never monitor through Bluetooth headphones or speakers. Bluetooth audio adds anywhere from tens to hundreds of milliseconds and no setting can remove it. Wired headphones only, plugged into the interface.
• Plug the interface directly into a port on the computer, not through an unpowered hub shared with other devices.
• On Windows, set the power plan to High Performance; aggressive power saving throttles the CPU mid-buffer and causes dropouts.
• Close browsers, chat apps, and anything syncing files in the background while tracking.
• Disable Wi-Fi and Bluetooth radios during critical sessions if you're chasing intermittent clicks — on some systems their drivers hog the CPU at exactly the wrong moment.

A note on sample rate: recording at 96 kHz halves the time each buffer represents, so it does reduce latency on paper. It also doubles the CPU and disk load, which often forces a larger buffer and cancels the benefit. It's an honest trade-off, not a cheat code — most home recordists are better served fixing drivers and buffers at 44.1 or 48 kHz.

Two Things Latency Is Not

Latency doesn't misplace your recordings. DAWs measure the driver's reported latency and shift recorded audio back into the correct position automatically. If your recorded takes consistently land late or early on the timeline, that's a compensation or driver-reporting issue — most DAWs have a recording-offset calibration for exactly this — not something you fix with buffer size.

Latency isn't fixed by turning up your headphones. Cranking the cue mix to "hear yourself better" doesn't reduce delay; it just risks your hearing, and hearing damage from sustained loud headphone monitoring is permanent. Keep monitoring at moderate levels — if you can't hear yourself clearly, fix the cue mix balance and your input levels instead. Healthy levels start at the source, which is a gain staging problem, not a volume-knob problem.

Why Zero Isn't the Goal

Even a perfectly configured system has a few milliseconds of round-trip latency — converters alone need time to do their job. Chasing zero is a waste of energy. Professional studios don't have zero latency either; they route monitoring through hardware paths (exactly what your interface's direct monitor knob does) and stop thinking about it.

That's the real fix: get the delay below the threshold where it affects your playing, by the cheapest and simplest route available, then forget about it and go record something.

The one-line version: monitor through your interface's direct path when you can; when you can't, track at 64–128 samples on a proper ASIO or Core Audio driver with a light session, then raise the buffer for mixing.