Exporting stems and "Normalize to Integrated Loudness"

When one exports stems from group channels, does the “Normalize to Integrated Loudness” work on those? I know it works on regular output channels, but I’m getting some overages on stems (such as mix-minus, M&E, etc.). It may be happening because most of the stems are undipped, therefore out of mix range and causing the red flags. Anybody else experiencing this?

Yes, I do think so.
That being said, I never tried this, because it simply confuses me.
When you need to deliver STEMS, there are no specs for them. The summ of the STEMS should match the final mix.
Or -when undipped MX is required- as close as possibly can match the final mix.
Sure, they can peak to -1dBTP if -xdBTP is required, but that is no problem at all; there are no specs for stems.
Except for the fact that they can’t clip.

If your STEMS are “over” (in terms of clipping) , then it must be that you are using some significant processing on your master buss.
I’m sure you have your reasons, but this sounds very strange to me.


OT question: What does the term “undipped” mean, in this context …? Thx! :sunglasses:

Depends on the spec (in the US).

For some it’s pretty simple, “undipped” means only to not dip levels for narration (or “vo” as some call it). So your “Mix Minus Narration” could be defined as “undipped” for example and that would mean that you would need to deliver one version where all elements are dipped wherever there is narration, and then another version without narration where levels are effectively flat where narration would have been.

But there are other networks that have other approaches to it, for example demanding that literally everything except the full mix is “undipped”, and not only just for narration but also for dialog. This means that your “Mix Minus” will contain everything with levels moving except where there would be narration, but your M&E stem doesn’t dip for either narration or dialog.

I personally detest the latter specs but it is what it is.

What Mattias said.

In simpler words: undipped Music STEM = MX without the fader-ride automation.


Thanks a bunch, guys! “Learn something new every day”. :sunglasses:

Undipped in this case (as most) is used for foreign versions. The network might have to lengthen or shorten translated sections, so cutting into music that’s undipped will yield a better result than if there is a fader movement in there.