Latency Buffer Calculator: Buffer Size to Milliseconds
Latency in milliseconds equals buffer size in samples divided by sample rate, times 1000. A 128 sample buffer at 44.1 kHz is 2.90 ms in one direction, and the round trip you actually feel is roughly 8 to 12 ms once you add the return path and 2 to 6 ms of converter and driver overhead.
Definition: Buffer size is the number of audio samples your computer collects before processing them as a block, and it trades delay against the risk of a dropout: a bigger buffer is safer and slower, a smaller buffer is faster and more fragile.
Every buffer setting is the same trade. A small buffer means your computer processes audio in small chunks, so the delay is short and the risk of a click or dropout is high. A large buffer means long chunks, comfortable processing headroom, and a delay you can feel in your hands. The arithmetic below tells you exactly where any setting lands.
What does the formula actually mean?
A digital audio system does not process one sample at a time. It waits until it has collected a block of samples, hands the whole block to the processor, and gets a processed block back. The buffer size is how many samples are in that block, and the time it takes to fill is simply the number of samples divided by how many samples arrive per second.
latency (ms) = buffer samples / sample rate x 1000
Worked through: 128 / 44100 x 1000 = 2.90 ms. At 48 kHz the same buffer is 2.67 ms, because samples are arriving slightly faster so 128 of them represent slightly less time. At 96 kHz it is 1.33 ms.
That is one direction. Your signal has to make the trip twice: in from the microphone, and back out to your headphones. So the buffer contribution to round trip is roughly double, which puts 128 samples at 44.1 kHz at about 5.80 ms before anything else is counted.
Why is the real number always higher than the DAW says?
Because the buffer is only one of five delays in the chain, and it is the only one you control. The others are fixed by the hardware:
- Analogue to digital conversion. The ADC uses a digital filter that has its own group delay, typically 0.5 to 1.5 ms.
- Bus transfer. Moving the block over USB or Thunderbolt takes real time. USB 2.0 class-compliant drivers are usually the worst offender here.
- Driver scheduling. The operating system decides when the audio thread runs. Core Audio and ASIO are good at this. WDM and generic drivers are not.
- Digital to analogue conversion. Same filter delay on the way back out.
- Safety offset. Many drivers add an extra block of padding to avoid dropouts, and some interfaces expose this as a setting.
Together these commonly total 2 to 6 ms. That is why an interface with an excellent driver, like the RME Babyface Pro FS , feels noticeably tighter than a budget box at the identical buffer setting even though the arithmetic on the buffer is the same. You are not buying lower buffer latency, you are buying less of everything else.
The practical habit that solved this for me permanently: two buffer settings, switched deliberately. Tracking sits at 128, and I accept that a heavy session will need some tracks frozen. Mixing sits at 1024, and I stop thinking about CPU entirely. What wastes time is leaving the buffer at one compromise value all day, because you get a session that is both slightly laggy to play and slightly prone to dropouts. Pick a number for each job and switch.
How much latency can you actually feel?
Sound travels at roughly 1130 feet per second, which works out to about one millisecond per foot. That gives you the single most useful intuition here: a round trip latency of 10 ms is the same delay as standing 10 feet from your amp. Guitarists do that constantly and think nothing of it. It is not the delay itself that is the problem, it is the delay combined with the fact that you also hear your acoustic sound with no delay at all, and the two arriving at different times is what feels wrong.
| Round trip | Equivalent distance | What it feels like | Suitable for |
|---|---|---|---|
| under 5 ms | under 5 ft | Indistinguishable from no delay | Anything, including live drum programming |
| 5 to 10 ms | 5 to 10 ft | Detectable but comfortable | Vocals, guitar, keys, most tracking |
| 10 to 15 ms | 10 to 15 ft | Noticeably behind on transients | Sustained parts, pads, soft vocals |
| 15 to 25 ms | 15 to 25 ft | Clearly wrong on anything percussive | Editing and mixing only |
| over 25 ms | over 25 ft | Unplayable | Mixing, mastering, playback |
What buffer should you use for each job?
Two settings, not one. Set the buffer for what you are doing right now and change it when the job changes.
- Tracking: 64 to 128 samples. At 44.1 kHz that is 1.45 to 2.90 ms one way. If your session is too heavy to run there, freeze or bounce the plugin-heavy tracks rather than raising the buffer. A performer playing to a delayed monitor gives you a worse take than any plugin will fix.
- Overdubbing into a busy mix: 128 to 256 samples. The compromise setting. Use direct monitoring if the interface offers it, which removes the buffer from the equation for the performer entirely.
- Mixing: 512 to 1024 samples. Latency is irrelevant when nobody is playing. Delay compensation handles the alignment and the extra headroom stops the dropouts.
- Mastering or a very heavy final mix: 1024 to 2048. At 2048 samples and 44.1 kHz you are at 46.4 ms one way, which would be unusable for playing and is completely fine for pressing space bar and listening.
What actually fixes latency, in order
- Turn on direct monitoring. Free, instant, and removes the problem entirely for the performer.
- Lower the buffer and freeze tracks. Costs you a few minutes and a little flexibility.
- Remove plugins from the input chain. A linear phase EQ or a lookahead limiter on the record path adds its own latency on top of everything above.
- Get a better interface driver. This is the real reason people say a particular brand feels tighter. Compare the options in our audio interface roundup.
- Raise the sample rate, carefully. Real gains at a fixed buffer, often cancelled out by the extra processing load.
Driver quality is the single biggest lever on real-world latency once you have set a sensible buffer, and it is the one thing you cannot fix in software after the purchase. Work through the full decision in the interface buying guide or the latency troubleshooting guide, and check the full conversion grid on the sample rate and latency chart.
Related tools
- Gain staging calculator: get the level right once the timing is right
- Interface spec database: connection type, preamp count and sample rate ceiling for every common model
- Signal flow explained: where in the chain each delay happens
- The intermediate studio build: the whole rig, priced
Frequently asked questions
How do you convert buffer size to milliseconds?
Divide the buffer size in samples by the sample rate in hertz, then multiply by 1000. A 128 sample buffer at 44,100 Hz gives 128 divided by 44100, which is 0.0029 seconds, or 2.90 milliseconds. That figure is one direction only. The round trip you actually feel is roughly double it, plus another 2 to 6 milliseconds of converter and driver overhead that no buffer setting can remove.
What buffer size should I use for recording?
Use 128 samples for tracking anything you play in time, which is about 2.90 ms one way at 44.1 kHz and typically 8 to 12 ms round trip on a decent interface. Drop to 64 if your machine can hold it and you are recording percussive parts. Then raise the buffer to 512 or 1024 for mixing, because latency stops mattering the moment you are no longer performing and the extra headroom lets plugins run without dropouts.
How much latency is too much?
Under about 5 ms round trip almost nobody notices. Between 5 and 10 ms most singers and guitarists start to feel that something is slightly behind them even if they cannot name it. Past 15 ms the timing error is obvious on anything percussive, and past 20 ms it becomes genuinely unplayable for drums or fast picking. Sustained pads and vocals tolerate more than transient-heavy sources do.
Why does my DAW report lower latency than I feel?
Because the DAW usually reports the buffer time in one direction only, and the round trip is at minimum double that. On top of the doubling you have the analogue to digital converter delay, the digital to analogue converter delay, the USB or Thunderbolt transfer time and the driver scheduling, which together commonly add 2 to 6 ms. The number in the preferences window is a component of the delay, not the delay.
Does a higher sample rate reduce latency?
Yes, at a fixed buffer size, and it is real. A 128 sample buffer is 2.90 ms at 44.1 kHz but only 1.33 ms at 96 kHz, because the same number of samples represents less time. The catch is that 96 kHz roughly doubles the processing load and the disk throughput, so many machines that hold 128 samples at 48 kHz cannot hold it at 96 kHz. You often end up back where you started.
What is direct monitoring and should I use it?
Direct monitoring routes the input signal to your headphones inside the interface hardware, before it reaches the computer, so the performer hears themselves with effectively zero delay regardless of buffer size. Use it whenever you are tracking without needing to hear a plugin on the input. The tradeoff is that you monitor the dry signal, so if a singer performs better hearing reverb you need either an interface with onboard effects or low enough buffer latency to monitor through the DAW.
Working out your own room and signal chain? The Home Studio Build Planner is the paid version of these pages: 8 printable worksheets you fill in with your own numbers, plus the full PDF, $29.