Free tool
Buffer size latency calculator
Turn a buffer size in samples into milliseconds at 44.1, 48, or 96 kHz, with the full preset table at your chosen rate. It is one formula: latency equals samples divided by sample rate. No mic, no audio, nothing leaves the page.
Buffer math
One buffer stage, in milliseconds
Latency = samples / sample rate x 1000. Pick a rate and a buffer, or type any buffer size, and the numbers update live.
Sample rate
Buffer size (samples)
One buffer, one way
10.7 ms
512 samples / 48000 Hz x 1000
x2 floor for a full path
21.3 ms
Input buffer + output buffer, before processing and driver overhead
These are calculated figures, not a measurement of your system. One buffer is one stage: a real monitoring path is at least an input buffer plus an output buffer plus processing, so double the one-way number for a floor estimate, and expect the measured figure to be higher.
Every preset at 48 kHz
| Buffer | One way | x2 floor |
|---|---|---|
| 32 | 0.67 ms | 1.3 ms |
| 64 | 1.3 ms | 2.7 ms |
| 128 | 2.7 ms | 5.3 ms |
| 256 | 5.3 ms | 10.7 ms |
| 512Current | 10.7 ms | 21.3 ms |
| 1024 | 21.3 ms | 42.7 ms |
| 2048 | 42.7 ms | 85.3 ms |
No audio is played or captured. This tool is arithmetic only, so nothing leaves the page and no permissions are requested.
If the buffer itself is the fuzzy part, samples, why audio travels in blocks at all, and what your driver does with them, start with what audio buffer size actually is, then come back and run your own numbers.
What the numbers mean
One way is the time one buffer stage adds: a 512-sample buffer at 48 kHz takes 512 divided by 48000 seconds to fill, which is 10.7 ms. Nothing downstream can touch the audio until the buffer is full, so that fill time is a delay you always pay, once per buffer stage.
The x2 floor exists because a real monitoring path is never one stage. At minimum the input side fills a buffer, your software processes it, and the output side fills another buffer before your headphones hear anything. Doubling the one-way figure is a floor estimate for a full path, and driver overhead, extra safety buffers, and converter delay stack on top. The only way to know your actual number is to measure it; this page is arithmetic, not a measurement of your system.
Raise or lower? Smaller buffers feel more immediate, and roughly 10 ms and under reads as immediate to most people. But a smaller buffer gives your CPU less time to fill each block, and a missed deadline is an audible crackle or pop. So the trade runs one way: if your audio crackles, raise the buffer or lighten the load, and see how to fix crackling and popping audio. If monitoring feels sluggish and your system has headroom, lower it one step at a time until it crackles, then back off one step.
Where the buffer lives: for an audio interface, the buffer size is set in the interface's own control panel, the ASIO driver settings, not inside most apps. patchd has no buffer picker at all: the engine locks onto the fastest path the device offers automatically and shows the live measured result in its read-only master clock readout, about 10.7 ms through the bus at a 512 buffer at 48 kHz, and smaller ASIO buffers take it lower.
See the measured number, not the math
This calculator estimates one buffer stage. patchd is a Windows audio mixer that measures the real figure on your hardware and shows it live in its master clock readout. Join the waitlist to be the first to know when it ships.