Volume drops after an hour

I just finished recording a novella of mine. It’s five hours total and I broke it up into five Audacity projects of about an hour each. When I look at the second section (chapters 11-20) and fit it all to the screen, it is obvious that the volume declines slowly from left to right. Using Analyze/contrast I see that the RMS in the first chapters is about -36 dB and in the last few chapters it is -43.If I listen to it and watch the bar graph, the early part peaks at about -20 and the later parts peak at -24. I was not conscious of speaking softer and I didn’t change any of the settings on my equipment. My noise floor is consistent across all ten chapters, about -80 dB. I’m using an AudioTechnica 2020 concenser mic into a Focusrite Scarlett interface, so I have a very quiet signal path.
Is this going to be a problem? If I normalize will it bring the last chapters up more than the first chapters and even everything out? Or will it normalize so that the louder chapters determine the amount of gain and the quieter chapters therefore don’t get enough gain and are still quieter than the earlier part? Can I normalize this hour long file in sections?
Also, have I been to cautious with my settings? Should I adjust my interface and computer settings so the bar graph will read higher, such as -6? Should the Analyze/contrast measurement be higher also? I’m still not sure why the Analyze contrast readings are so far from the bar graph readings. When I examine the same section of audio the bar graph shows -20 dB but the Analyze/contrast measures -37.

I just finished recording a novella of mine.

With the goal of posting it on ACX AudioBook, or other company?

Which computer and Audacity are you using?

The blue waves on the timeline will give you a measurement of peaks or instantaneous vibrations of the air. That’s a good measurement of what the audio data is doing. It’s in percent.

Analyze > Contrast uses RMS measurement. That’s a complex math measurement of the area under the curve for each blue wave. It so happens to work out to be a very good measurement of loudness, almost always a lower number than peak.

The third measurement is Noise, the loudness of the show when you stop talking. So those are the three AudioBook standard measurements.

It’s Most Unusual for a presentation to slowly fade into the sunset as yours does. I don’t know that I can bring that home to a defect or other bad adjustment. What you’re supposed to do is watch, out the corner of your eye, the bouncing light Audacity sound meters as you work. That’s why the new sound meters are very much larger and green-yellow-red multi-colored compared to the older ones.

If you’re self-recording, you are taking the place of the audio engineer who would normally be watching the sound meters for you.

Effect > Normalize is badly named. It suggests you run the tool to even out all the volume variations and we can all go home.

No. All it does is turn the volume up or down to an entire selection at once. If you have a chapter whose beginning is twice as loud as the end (6dB), it will still be twice as loud when Normalize gets done. Normalize and Amplify are close sisters. They both just turn the volume up or down once, but you tell them how to do it differently.

So are you hoping for an ACX posting? We have special tools that can help including a measurement tool that folds all three ACX measurements into one panel.

We would rather posters settle their recording system and technologies before they finish reading a whole book. Is this your first book? Did you find your reading quality and presentation improved through the course of the book? Did you get the urge to re-read the first few chapters when you got done with the first complete pass?


This is my first audiobook recording but some years ago I worked as an announcer in a radio station and I’ve also had singing lessons, so I’m aware of the need to breathe deeply and project my voice to keep up the volume so it doesn’t drop at the end of the sentence. I am preparing this book for Audible.com. So far I’m working in WAV. I sent ACX a short excerpt and they approved it, so I thought that I was ready to go. They don’t recommend noise reduction so I didn’t use it. My signal chain is very quiet so I believe I can meet their standard without NR. All I used on that excerpt was normalize, -3 dB limiter and filters below 100 Hz and above 16K.
I’m reading off my laptop while recording, so I can’t see the lights on Audacity when I’m speaking. I check by reading a few sentences and watching the bar graph each day but then I have to switch the screen to the novel manuscript while I’m reading. I realize that I should use a split screen or get a bigger monitor so I can have Audacity and my script both in full screen size. Even then, however, I can’t watch two screens at the same time and my concentration will be on the text.
I thought that I had the latest version of Audacity but my version doesn’t look like the screenshot you enclosed. I like the longer bar graphs. I will check and upgrade to the latest version. I did notice that when I finished and went back to listen to the first chapters, there were some places that I wanted to read differently, but not all. Some of the early chapters are fine, assuming that the recording levels are usable. I’m using a Dell laptop with Windows 8.1.
From what you say about normalize, I can select any portion of a chapter and normalize it separately from the rest. Assuming that I do this intelligently, the average volume of all chapters will be the same, but will the parts that are now six dB lower sound too much different from the parts that are raised less? Will the tone of the spoken parts be different? And the spaces between the words might sound different because I’ll be amplifying the room tone also. I have a very low noise floor so I expect that will help. Thanks for your help.

I was running Audacity 2.1.2 but the screenshot you sent doesn’t match what I have on screen so I updated anyway and I still have a different screen. The recording volume and speaker volume bar graphs are both on the same line and only go from -60 to 0. Is this a difference between Mac and Windows? Can I reconfigure the tool bar?

I sent ACX a short excerpt and they approved it, so I thought that I was ready to go.

That would be our assumption, yes.

They don’t recommend noise reduction so I didn’t use it.

Two comments: The earlier Noise Removal sucked, so using it was a major effort and most people use too much. The object is gently lower noise, not produce The Deep Blackness of Space between your words. This is the first place having an old Audacity is not desirable.

I used on that excerpt was normalize, -3 dB limiter and filters below 100 Hz and above 16K.

We have a pre-baked filter for that 100Hz thing now. You should be here giving advice to others.

I can’t see the lights on Audacity when I’m speaking.

I can (attached.) I’m supposed to be able to detach the meters, but that doesn’t work.

As we go

AudacityMetersAnd Script.jpg

Turns out I can place the meters on top of the Audacity window taking up even less real estate and I have always been able to use either vertical or horizontal meters. The floating meter thing seems to be broken.

As we go.

Screen Shot 2016-03-15 at 11.42.20.png

Audacity > Edit> Preferences > Tracks > Meter Range…

The beginning and end bars (left and right) are position and configuration tools. You can push all the bars around to do what you want.

I thought the default was one over the other…

Did you get your new Audacity 2.1.2 from here (I’m assuming Windows).



I got my Audacity from the Audacity website. I can move the bar graphs around by using the cross hatched vertical panel on the left side of each bar. I can widen and deepen them using the handle in the lower right corner, but the lowest reading remains -54. When I go to Edit/Preferences/Tracks there is no “meter range” choice available.

Sorry. It’s Preferences > Interface.

-60 is not a dreadful meter setting for recording because the majority of your work will be in the top 40dB or do. Playing, editing and production is different because of the need to manage noise which should be lower than -60—and you need to be able to see that accurately.

If you keep expanding the left meter with the right-edge handle, eventually the two meters will flip one over the other—because they’ll have nowhere else to go.


I like my meters @-120dB, so I can see the noise floor and the real dynamic range my gear gives me at that precise moment.

It shows when I’m in a location that has particular noise…