Why does RMS change for Mix and Render when only one track?

When I mix and render a single track, the contrast analysis RMS can change. Method:

  1. record 10 secs voice to track A
  2. mute track A
  3. record 10 secs voice to track B
  4. copy some of track A, paste into track B
  5. measure RMS of track B
  6. Mix and Render track B to new track C
  7. measure RMS of track C - it’s different from #5 reading

If I invert track C, then mix and render B & C to track D, I get a flat line, so the reported value in #5 is off. If I analyse amplitude statistics, I get identical values for each track, matching the reading for track C. I infer from all of this that contrast analysis readings are approximate.

The reason I’m checking this is to set correct RMS levels for ACX audio. I can live with it as long as I understand what’s going on.

v2.1.2 .exe installer on Windows 10.

There’s a tool that put all the critical ACX measurements in one INFO panel.


Download it and install it as a plugin. Analyze > ACX Check

There’s no contest between that and struggling step-by-step through Contrast and Amplify. This is what that looks like if you haven’t beat your way through the whole thing yet. I think this will only do a mono show, not stereo.


There’s some interesting tricks to doing this reading thing. If you feel like it, submit a voice test.


A very common mistake is using too much processing.


I’m not able to reproduce the issue. Is it repeatable for you?
If so, could you make a ZIP archive of the project (including the AUP file and its _data folder) and upload it somewhere, then post a link to it. There’s some suggestions of free file sharing sites near the bottom of this post: https://forum.audacityteam.org/t/how-to-post-an-audio-sample/29851/1

I cannot reproduce it either, but why is pasting part of A into B a necessary part of the steps?

I would guess an explanation could be that you have Transport > Overdub on, and some of track B is thus pushed behind zero, and that you then time shifted track B or C. It seems Contrast does not measure audio behind zero, but Amplify for example does.


Took me a few attempts to reproduce. I suspect it may require the pasted clip to be lower level than the destination. Project is at http://www.philmayes.com/audio/RMS.zip

I duplicated track B before the paste to make it a little clearer. For the 3rd & 4th tracks, I get contrast analysis RMS of -28.0 dB and -27.5 dB. Yell if you need more info or tests.

Thanks Phil. Yes it seems even duplicating track A (or B), copy and paste from one duplicate to the other then Mix and Render to New Track can produce different RMS between original and rendered track.

We already had a bug open for Contrast measuring identical tracks differently, which I have updated with your findings: http://bugzilla.audacityteam.org/show_bug.cgi?id=1320.


I tested this bug today with Phil’s Steps to reproduce in his original post - testing on macOS High Sierra and on W10 home with the latest 2.2.1 alpha nightlies.

On both platforms the RMS measurements remained the same in the rendered track as in the original track.

Accordingly I have closed this bug as RESOLVED FIXED: http://bugzilla.audacityteam.org/show_bug.cgi?id=1320

Thanks for the report Phil.


I’m checking this is to set correct RMS levels for ACX audio.

There is a suite of tools for producing ACX audiobook readings. You don’t need to go through that Contrast thing any more and you don’t have to cycle through compressors or other gymnastics to set RMS levels.



Thanks Koz. I knew about acx-check, but not rms-normalize.

I knew about acx-check, but not rms-normalize.

If the goal is submission to ACX, the tool collection is indicated. The steps were tested as a suite and there is a warning to take them in order and not leave anything out. Each tool depends on the results of the tools that came before.

For example, the first effect, Low Rolloff is a rumble filter and gets rid of very common sub-sonic trash that will throw off RMS-Normalize and the Limiter.

Home recordings commonly suffer from noise and we decided to push that off to its own document. If you have difficult noise, no tools are going to be easy and the condition may be a showstopper. Noise Reduction may plain fail, or too much Noise Reduction will fail ACX Human Quality Control. Some common microphone noises have specialized repairs. If you find yourself going round and round, post a sample and we can give a diagnosis.



As it says several times in the paper, the tools only work in Audacity 2.1.3 or later.