Questions about details of plug-in latency

When viewing the plug-in manager there is an information pane at the bottom which displays the latency for each plug-in.

What is that number a measurement of? Samples?

Does latency stack (in serial?) from multiple instances of the same plug-in, or does the total latency remain the same with multiple instances (in parallel?)?

Put another way, does the total latency added by plug-ins come from adding all of the latencies together (including multiples for each instance), or is the total latency represented by the plug-in with the highest latency measurement?

The latency is displayed in milliseconds.
It’s cumulative, they all add up including multiple instances of the same plugin.

Your answer is certainly not correct.

When I am running 30 instances of a plugin that has a latency rating of 4090,
by your account I should have over 122 seconds of latency (4090 x 30 = 122700, 122700ms = 122.7 seconds). I don’t. The RTL I experience is well under 1 second.

Anyone else?

Are all the plugins on the same track?

I think you misunderstand. GFY.

planarchist - “I think you misunderstand. GFY.”

If am wrong, it is easy for you to show how. I laid out the math which is a direct consequence of the answer you gave. Go check the latency ratings of your own plugins in the plugin manager, do the math of the formula implied by your explanation (x milliseconds • n instances), add up 20 instances of a few plugins and see if you get an accurate latency time from that.

Even more to the point is taking just the Steinberg DeEsser as an example.
The latency rating on that plugin is 1624. If you are correct, and the latency here is measured in ms (milliseconds) then one instance of the DeEsser would cause a whopping 1.624 seconds in latency to use. That is certainly not the case!

This is likely to be samples not ms.
(You can check the channel latency in the mix console)

Think about the signal path. If you put several insert effects on one track then those effects are processed one after another and the latency of each plugin adds up. Take into account any groups and master bus effects.

1 Like

Yes, samples would make the most sense I think.

Now about serial vs parallel-
I understand that serial (say stacking the same plugin as an insert) would cause cumulative latency (given the signal flow must be linearly additive), but what about parallel instances? If you have 10 compressors running on 10 separate channels, would the latency remain the same as if you are using one instance?


Cubase uses delay compensation to keep the project audio in sync, so the channel with the highest latency would determine the latency of the project audio.

If you are only running those 10 compressors and they introduce an identical latency per channel, the project latency would remain the same as if you were using a single instance.

At some point your CPU might however run out of processing power at the current sample buffer size, resulting in crackles in the audio. To get around this you’d either have to deactivate plug-ins, or increase the sample buffer size which in turn increases the total latency.