Buffer Size vs. Audio Latency Explained

Quick Answer

Audio latency is usually a buffer-and-processing problem, not a mystery number. The buffer formula gives you the starting point, but the real-world result depends on drivers, plugins, conversion stages, and whether you are measuring one-way delay or full round-trip latency.

What is Audio Latency?

Audio latency is the time delay between an audio signal entering a system and emerging from it. In broadcasting, live events, or studio recording, this delay accumulates as the audio travels through ADC converters, computer processing buffers, digital network protocols, and finally the DAC interface.

Whether the path uses Dante or AES67, a software mixer, or a local audio interface, separate the theoretical buffer duration from the measured end-to-end result.

How to Calculate Audio Latency

Buffer size and sample rate define the duration of one buffer. They do not by themselves define the full latency of OBS, vMix, a DAW, an interface, or a networked audio path.

To eliminate manual calculations in the heat of a production, use our audio buffer latency calculation tool. It gives you the calculated delay for common sample-rate and buffer combinations, which you can then verify against the actual signal path.

Examples: 128, 256, and 512 Samples

Here are the theoretical single-buffer durations for common buffer sizes at 48 kHz; driver, safety buffer, conversion, and application stages add more delay:

• 128 samples = 2.67 ms: one-buffer duration; lower settings generally increase scheduling and CPU pressure.
• 256 samples = 5.33 ms: one-buffer duration; suitability depends on the complete monitoring and processing path.
• 512 samples = 10.67 ms: one-buffer duration before conversion, safety buffers, plugins, and output buffering.
• 1024 samples = 21.33 ms: one-buffer duration; often chosen when processing stability matters more than interactive monitoring.

Driver Architecture: ASIO vs. WDM

On Windows-based production systems, the available driver path can materially affect latency and stability, but it is one part of the complete path:

• WDM / WASAPI: The standard Windows driver path. It often adds more latency and less predictable behavior than an ASIO workflow, though the exact result depends on the device and software stack.
• ASIO (Audio Stream Input/Output): A direct driver model commonly used for lower-latency professional work. It is often the better choice for sub-10 ms targets on Windows, but the result still depends on hardware, plugins, and routing.

Total Round-Trip Latency (RTL)

Buffer latency is only one stage. Audio round-trip latency can include:

1. A/D Conversion: device-dependent input conversion
2. Input Buffer: (e.g., 128 samples @ 48kHz = 2.67ms)
3. Processing: VST plugins, EQ, Compression (variable)
4. Output Buffer: (e.g., 128 samples @ 48kHz = 2.67ms)
5. D/A Conversion: device-dependent output conversion

The driver-reported round-trip value and a loopback measurement may differ because safety buffers, conversion, USB or network transport, and plugins are not always represented the same way. Measure the real path when the number matters.

Common Mistakes

• Treating buffer math like the full end-to-end latency answer.
• Ignoring plugin delay and only checking the interface buffer size.
• Confusing one-way latency with round-trip latency.
• Assuming one Windows driver path behaves like another.

References

• OBS Knowledge Base
• Audinate Learning
• Steinberg ASIO Overview