Editing in 32-bit vs 16-bit sampling formats

So, I have a FLAC 16-bit music track that I plan on adding some repeated sections to and eventually export it as FLAC 16-bit again. I was wondering if it would be better to edit this in 16-bit default sample format or 32-bit.

From what I understand, if I use the 32-bit format, I would later have to apply some dithering when exporting to 16-bit to avoid errors. But if I use 16-bit format, would I be able to skip dithering? And is editing in 16-bit format even a good idea in general? Most of the edits I’ll be doing are copying, time shifting, mixing of overlapping tracks, and maybe fade in/out.

Also, I’m running Audacity 2.1.3 on Windows 10 x64.

It’s not 32. It’s 32-floating. Audacity uses that internally to avoid processing damage during editing.

32-floating doesn’t overload. If you apply a filter or effect that causes your show blue waves to rise over zero—to go into clipping, you can just bring the volume down later and keep going. In the other fixed-point formats, that may be the end of the show.

This protection is only available inside Audacity. The microphone before Audacity and normal sound files after Audacity still overload like they always did.

I believe Audacity uses shaped dithering default now. You can’t hear it working in normal presentations. If dithering becomes objectionable, you may have more serious problems than that.


Thanks for the info. I’ll probably just stick to editing in 32-bit float format.

Audacity processes internally at 32-bit float even if you choose 16-bit quality. If the processing involves changing the sample amplitudes, as mixing and fades do, dithering would thus occur when the 32-bit processed audio is returned to the 16-bit track.

If you were just doing deletes and repeats, you could avoid dithering and stay in 16-bit, but you would still have to turn dithering off in Quality Preferences because there is a bug that adds un-necessary dither when you export from a 16-bit project to a 16-bit audio file.