Dithering is a tiny amount of specially shaped noise added on purpose when you reduce the bit depth of audio, for example when you export a 24 bit mix down to a 16 bit file. It sounds backwards to add noise to clean up a file, but that small noise replaces a harsher problem: the rough distortion that appears on quiet fades and tails when the finer detail is thrown away.
This is a plain explanation of what dither does, why it helps, and the one moment you actually reach for it. In most DAWs it is a single checkbox in the export dialog, so in practice it is a set and forget step at the very end, not something to agonise over.
Dithering is a tiny amount of specially shaped noise added when you reduce the bit depth of audio. The everyday case is exporting a mix you recorded and worked on at 24 bit down to a 16 bit file for release. Bit depth is how precisely each sample is stored, so dropping from 24 bit to 16 bit means keeping fewer of those fine steps and rounding the rest.
When you round every value to a coarser scale like that, you introduce small errors. Dither is low level noise added just before the rounding, so those errors are scattered randomly instead of following the music. The result is that the tiny mistakes turn into a faint, steady hiss rather than a harsh, grainy distortion. You are trading a problem your ear catches for one it mostly ignores.
In most DAWs this is a single option in the export or bounce dialog. You switch it on when you render the final 16 bit file and leave it alone otherwise, which is why it is fair to call it a set and forget step. The rest of this guide is really about why that checkbox exists and the one time you use it.
You dither because chopping off the lower bits on its own creates an ugly distortion, and the dither noise smooths it into something far less noticeable. Without dither, the rounding error is tied to the signal itself, so on quiet passages it shows up as a gritty, buzzy distortion that rises and falls with the music. That is exactly the kind of error the ear is good at catching.
It is worst where you would least want it: quiet fades, the tails of reverbs and delays, soft intros and endings. As the sound drops toward silence, fewer and fewer bits are describing it, so the rounding gets relatively larger and the distortion becomes more obvious. A clean fade can end up sounding crunchy or granular right as it disappears.
Adding dither randomises that error so it no longer follows the signal. Instead of distortion locked to the music, you get a low, smooth hiss sitting underneath everything at a very quiet level. Your ear treats steady background noise as far less objectionable than distortion that moves with the notes, so the file sounds cleaner even though, strictly speaking, you added noise to it.
Dither swaps a harsh distortion your ear notices for a low, smooth hiss it mostly does not. On music at a normal level that hiss sits far below what you can hear, and it only ever matters at the quietest moments, which is precisely where the untreated distortion would have been most obvious.
Dither once, at the very last step, and only when you are actually reducing bit depth, for example a final render to 16 bit. That final export from your higher bit depth working file down to the 16 bit delivery file is the one place it belongs.
If you are staying at 24 bit or higher, you do not dither, because nothing is being reduced and there is no rounding error to cover. For the same reason, an intermediate bounce that you are going to keep editing does not get dithered; you save it for the actual final file. If you want the background on why 24 bit is the working format and 16 bit is the usual delivery standard, that is in the guide on sample rate and bit depth.
And you never dither the same audio twice. Each pass adds its own layer of noise, so dithering an already dithered file, or dithering and then converting again, just stacks hiss for no reason. Add it a single time, on the final export, and stop there.
| What you are exporting | Dither? | Why |
|---|---|---|
| 24 bit mix down to a 16 bit file | Yes | The bit depth is dropping, so dither this final render. |
| A final file staying at 24 bit | No | Nothing is being reduced, so there is no rounding error to cover. |
| A 32 bit float working file | No | The working format keeps the full detail, so there is nothing to dither yet. |
| An intermediate bounce you will keep editing | No | Save dither for the actual final file, not a stage you will process again. |
| A file that was already dithered | No | Never dither twice. Each pass adds its own noise for no gain. |
Noise shaping is an option on many dither settings that moves the dither noise into the frequencies your ears are least sensitive to, usually the extreme high end. The total amount of noise stays about the same, but by shifting it to where your hearing is weakest, it becomes even harder to notice.
Human hearing is most sensitive through the middle of the range, roughly where speech and the body of most instruments sit, and much less sensitive at the very top. Noise shaping takes advantage of that by pushing the dither energy up out of the sensitive middle and into the high frequencies you barely register. It is the same idea as tucking a small sound where you are least likely to hear it.
DAWs offer different types or strengths of noise shaping, with names that vary from one to the next. For most music you can leave it on the default, or use plain dither without shaping, and be perfectly fine. Plain dither already does the important job of removing the distortion; noise shaping is a refinement on top, not something to lose time choosing between.
Dither is a small, deliberate noise added when you lower bit depth, so quiet parts stay smooth instead of turning grainy. Use it once, at the final export to 16 bit, and never twice. If you are not reducing bit depth, you do not need it. Noise shaping is an optional extra that hides the noise a little better, and the default setting is fine for almost everyone.
Dither is a single step at the very end, and it does its job quietly whatever you feed it. It sits alongside other final choices like how loud your master should be, but what really decides how good the file sounds is everything before it: the writing, the playing, the mix. That is where your attention is worth spending.
Our instruments make sound inside the computer, so they hand you a clean, quiet signal with a wide dynamic range and no tape hiss or room noise baked in. That means your mix arrives at the dithering stage in good shape, with real quiet parts for the dither to protect rather than a noisy floor already sitting under everything.
When you want them in a project, The Collection ships as AU, VST3, AAX and a standalone app on macOS, signed and notarized, and a VST3 on Windows, so it loads into whatever session and settings you already have.
Free to hear in the browser, no signup. Dither happens later, at export, not while you play.
We write these when there is something worth writing down. One email when a new one lands or a new Tunary instrument ships. No newsletter, no schedule.
Tunary