What Is Dithering in Audio?

Truncating bit depth without dither doesn't add hiss, it adds distortion that tracks the music. Dither converts that distortion back into honest, unobtrusive noise.

The reason the noise is worth it

How dither fixes it

Dither adds a tiny random noise to the signal before the rounding happens. That randomness decouples the rounding error from the music, so instead of a distortion that follows the notes, you get a steady, low-level hiss that sits politely in the background. The remarkable part: with dither, signals quieter than a single 16-bit step can still be reproduced on average. You're trading a structured, audible distortion for an unstructured, barely-audible noise floor, and the human ear vastly prefers the latter.

Most mastering tools go one step further with noise shaping: they push the dither noise up into the 15 to 20 kHz region where your hearing is least sensitive, making it even harder to notice. That's why dither plugins often offer several 'noise shaping' curves, they're all the same idea, tuned for different material.

When to dither, and when not to

This is where most mistakes happen. The rule is simple: dither only when you reduce bit depth, and only once, at the very end.

The single most common error is dithering more than once, for example dithering a stem, then dithering again on the final master. Each pass adds another layer of noise for no benefit. Dither is the last thing that touches the audio, full stop. Anything after it (even a gain change) technically undoes the careful rounding, which is why it goes dead last in the chain.

Does dithering still matter for streaming?

Partly. If you deliver a 24-bit master to a distributor or platform, you don't dither at all, you hand off full bit depth and let their pipeline handle conversion. Dither matters when you are the one creating the final 16-bit file: a 16-bit WAV master, a CD, or a download you're selling direct. In practice, the safest modern workflow is to keep everything 24-bit through mixing and mastering and only dither on the specific deliverable that has to be 16-bit.

It's worth being clear-eyed about scale. Dither is a refinement, not a make-or-break, the difference lives in the bottom few decibels of the quietest passages. Get your LUFS targets, your balances and your dynamic range right first; dither is the polish you apply once everything else is finished.

Dither and file format: don't undo your own work

Here's the catch that quietly ruins careful dithering: it only survives in a lossless container. Dither lives in the lowest bits of the file; a lossy MP3 or AAC encoder throws those exact bits away and replaces them with its own quantization noise. So you dither to a 16-bit WAV or FLAC, not to an MP3. And once you've made that pristine 16-bit master, getting it to its destination unaltered is the whole game, drop it into a delivery Room rather than a chat attachment that will helpfully re-compress it for you.