There is still some tiny hum in there (below threshold). Sometimes you can get improvement by rotating the power supply while listening carefully with the volume up in your headphones. I got myself out of hot water once by finding the one place in the room where the microphone didn’t pick up hum… and did the announcing there.
Do you have any lamp dimmers? Didn’t we go through this list before?
I’m being obsessive. The noise isn’t obvious and I have to actually look for it. You should probably forget it. Marginal noises are only important if you have to do very serious filtering, compression, limiting or other process that changes the size of the blue waves. Then you start to worry. You got to ACX with your correction list.
I’ve been working on new files which Jerreth has begun to record. I do have just one question. I have been following Koz’s instructions. But I see no mention of using noise reduction. I just made it, barely, when I measure noise, without using noise reduction. But the noise level is MUCH lower when I do noise reduction in addition to the instructions. I have this nagging doubt, however, that perhaps this makes it too quiet. Here’s a sample of a file in which I used noise reduction.
Procedure followed:
Effect > Normalize > [X]Normalize to -3.2 [X]Remove DC
Effect > Compressor:
Threshold -20
Noise -50
Ratio 2:1
Attack 0.2
Release 1
Effect > Normalize (again) > [X]Normalize to -3.2 [X]Remove DC
Compress to 2:1
Normalize again
Equalize using LF Roll off for speech (Steve’s filter)
Then finally: Noise reduction having obtained a noise profile, first, from the beginning, where Jerreth records a few seconds of silence before reading.
This is where we bounce into technical versus theatrical. You can totally create a sound performance that passes ACX and doesn’t sound all that good. I did it recently where I was testing a microphone and I happened to create a performance that passes ACX conformance, but nobody would mistake me for being either entertaining or theatrically valuable.
I found that The Noise Reduction of the Beast can help to suppress just enough hiss and background to make the show pleasant to listen to. You are required to be in Audacity 2.1.0 for this. The earlier Noise Removal tool won’t do this.
Select the profile from your Room Tone quiet segment, and then apply Noise Reduction to the whole show with 6,6,6, settings. That should be enough to improve the quality without being obvious what you’re doing. Keep in the back of your mind that one way to fail ACX is “Overprocessing.”
I’m going back to look at your list again. I’ve been using LF_rolloff earlier in the process to keep rumble and other accidental very low pitch sounds from affecting the other tools.
This is also where we find out how obsessive you are. Are you keeping the raw performances in WAV? We urge strongly creating the final performance in WAV and then export it to MP3 for ACX submission, but only the true, card-carrying obsessives will also save the original performance. In movies this is called the “Camera Negative.” The film that actually went through the camera (in case anybody is still using film). "I have the Camera Neg, " is usually spoken with hushed, reverent tones. I still have both corrected and original shoot sound files from several years ago.
You might want to apply a different correction suite. You can’t take adjustments out of a corrected performance. You have to go back.
Also prominent in the ACX videos is the insistence that your chapters be absolutely consistent one to the other.
I hope that’s a typo. You claim you’re compressing twice and there’s no limiter in the list. I don’t think I’ve ever recommended two passes through any processor. Normalize is a volume change, not a processor.
Yes, that was a typo. I was using normalize on either end of compression.
Yes, I am saving the originals before I do anything to them. I am saving all my work with ID and date and when I try something new I also save the older work (at least until we’re done). I’ve been saving Jerreth’s initial recording as a Audacity project file. Is there any quality difference between this and a WAV file?
I’ll go back and do it with the 6,6,6, settings for noise removal.
However, it still makes me nervous your steps seem to be out of order. The order is important. Having Noise Reduction as the last step could easily mess up the peak or volume setting in the final show. Having LF_rolloff where it is could misdirect the compressor.
Wow! That was a thrill Koz! I thought, "Oh no! What have I done now?! But I checked and the original file I had worked on; the whole preface, measures 3.2 on peaks, when I select the whole file, click on amplify and read the number in the top box. I assumed that the measurements would remain the same when I selected a small part of that file and exported it, giving the test clip a new name. I didn’t go back and do the measurements on that shortened clip, assuming that it would have the same measurements as the larger file. But no, it doesn’t! When I did the peaks measurement on the shortened clip, I too, get 4.7. I don’t know why the qualities of the exported portion would have changed.
I’m going to export a shortened clip from the original Audacity project file and then the the adjustments on that. That should produce the right measurements.
Here it is. My measurements on this test clip are:
Well, I don’ t know what to think. I downloaded my own test clip and these are the measurements I got from it. It appears to have changed in the process of being uploaded or downloaded.
That’s just knowing what the tool is doing. At least once, somewhere in the hour long show, one peak made it out to 3.2. That counts.
When you supply a test clip, you should analyze that one clip and quote the findings. For the sake of argument, suppose you supplied a segment of the performance that had no silent stretches. It would drive the silence measuring tools nuts, but the rest of the show would probably be fine.
RMS Loudness, the middle number is more or less averaged over the show. That, too, can give nutso readings if you pick and choose a short segment of the show. If Jerreth was particularly expressive and somber right there, that one short segment will not pass ACX, but again, on average, the show is probably fine.
Even more slippery, we’re guessing that’s how ACX measures things based on our short conversations and actual performer’s successes and failures.
Sorry for cluttering the thread. Here’s one more attempt. I extracted my shortened clip, first, from the original aup file and followed your directions. This is what I measured Noise: -69.6
RMS: -20.0
Peaks: 3.2
Then I exported the wav file, clicking on “Export Audio” and keeping the same name, so as to overwrite the file with the corrected one. Finally I went through and did the measurements again, on the saved file.
Noise: -69.5
RMS: -20.0
Peaks: 3.2
So nothing has changed. now I’ll upload this file and test it, after downloading.
I just downloaded the file and tested it. Here’s what I came up with.
Noise: -69.4
RMS: -20.0
Peaks: 3.2
I really don’ t understand what happened. But I guess, when I want to do a test clip, I need to extract it from the raw project file and work on it as a separate file.