exporting to MP3

I just got a new mic and I’m also taking an audiobook class, and maybe my settings have changed, but here’s my question:

Why when I export to MP3 from my AUP file do the specs change slightly? This never happened before. For example, I run through ACX check on my AUP file, and it passes. And then I run ACX check on the exported MP3, and it still passes, but the specs have changed slightly. They are not the same as the saved file. This didn’t seem to happen before. Why is this happening and is there any way of preventing this from happening so that the saved file in audacity is the same as the exported file. I appreciate some insight into this issue. Thank you.

MP3 is lossy compression. Data is thrown-away to make a smaller file so when it’s decompressed you don’t get the original data back. You might see a change (usually an increase) in the peak level.

1 Like

Are you using our Audiobook Mastering process? Mastering uses a peak target of -3.5dB—slightly quieter than required—to make up for the possible increased distortion of the MP3 conversion.

ACX needs your submitted MP3 file to be 192 bitrate Constant.

Screen Shot 2023-02-25 at 13.10.11.png
They soft recommend that your submission be mono, but they will accept stereo. Whatever you pick, your whole book has to be that way.

192 Constant Mono is a very nigh quality standard and I would expect your Edit Master WAV file to closely match the MP3. You may be making the MP3 file wrong.

You are making a WAV Archive Edit Master of the chapters, right? You can’t modify or change an MP3 without causing sound damage.

Koz

You can use Audacity Projects (AUP) as masters, but Projects can be brittle and easily damaged. Projects only open in Audacity.

There are two different Project formats now. The new format does not use the split file process. New Audacity should open up both formats, but the older Audacity will only open the older format files.

WAV files are uncompressed, perfect quality, and hard to break. They are well understood and open up anywhere.

Koz

Yes, I know what ACX needs. I do use 3.5, though my teacher said ACX demands 3.0 peak though I thought it was a range and no higher than 3.0. I only do mono. Yes, I always save a WAV. That’s the issue. The MP3 is slightly different specs than the WAV.

You are correct.

ACX Website

RMS between -18 and -23dB.

Peaks no higher than -3dB. (And of course, since it’s a negative number -3.5 is lower and OK.)

(I don’t think it’s possible to hit BOTH numbers without limiting or compression.)

They give you a maximum noise level and they say there should be “room tone” but they don’t specify the minimum. They will reject you for “pure digital silence” which is -infinity dB, if it’s unrealistically low from inserting silence or “over processing” with a noise gate, etc.

P.S.
The difference between the WAV and MP3 should be repeatable for a given file so if you want it “exact” you can go back to the WAV and adjust the volume a fraction of a dB.

Presumably, you are somewhere in the middle of the RMS range and so slight adjustment shouldn’t throw that out of spec.

Because MP3 is a compressed approximation of the original sound file. MP3 gets its tiny, convenient sound files by carefully shuffling musical tones—and leaving some of them out. That’s why you never want to do serious production in MP3.

Having said that, ACX requires delivered MP3 work to be at 192 Constant quality and they strongly request you do the work in mono (one sound channel) and not stereo. If you do that, I would be surprised if you could find any difference between your Edit Master chapter and the MP3 submission.

When you get finished editing a chapter, Export a Perfect Quality WAV file and use that as your Edit Master Archive. You can save an Audacity Project if you want, but WAV files are robust, excellent quality, and they will open anywhere.

Then make the MP3 for ACX submission.

Koz

Which specs changed slightly?

Koz

I am referring to the peaks here. When I run my ACX check in the wav file, which, by the way I do create first and save, the normalization is set to -3.0 but when I export the MP3, it’s something like -3.27. It’s not exactly the same. But if it’s still OK for ACX, and I’m not going to worry about it. I guess I just wanted to know why that happened. But I do follow the instructions that you suggested in your response.

As a terrible example of how MP3 works: Say you converted an orchestra to MP3. The MP3 software discovered that they could get a slightly smaller file if they played the flutes the smallest fraction of a second ahead of everybody else. Nobody will notice that, but their notes happen to line up exactly with the oboes now and the overall size of the waves goes up.

MP3 takes this process down to the individual tones in your voice.

If you want to see (hear) this in action, take a short, throw-away sound file and compress it a couple of times at 32. This is what it sounds like when MP3 has to throw away too much quality.

Koz

thank you! It’s all a little bit too technical for me but I am guessing what you’re saying is the change is so minute that nobody would really notice it.
I am not really an audio engineer, but a narrator. So a lot of this is over my head. All I know is that I want to pass a ACX!

Are you using both Audiobook Mastering and ACX-Check? The only restriction to Mastering is the urging that you use the tools in order, don’t add any, or leave any out.

They clean up after each other.

It’s strongly recommended that you build a passable “studio.” You can patch a bad recording done with background noise, echoes, sibilances, popping, and other damage, but it takes time and you have to remember to do it all at every recording. Plus, you have to not overdo it. ACX has a failure called “Overprocessing.”

It doesn’t have to be crazy, either. I produced a very nice voice recording on my phone in the garage after hours.

Koz

i’m all set with what you were mentioning. I have a book on audible and it passed the first time around. I use ACX check and I have my macro set up for mastering. but I also have a program called Levelator which does it as well. i’m pretty good editing out little noises. And I only use noise reduction if I have to.