Audacity randomly recording at different volumes?

Using Audacity 2.1.3 on my 8-year-old Dell PC running Windows 10. I’m using a Blue Snowball mic.

I had recorded a narration for a podcast, and I noticed that when I amplified the track to my normal multiplier (8), it clipped everywhere. So I undid it and realized that it had magically recorded my narration at an 8x amplification. I thought it was strange, but no harm done I guess. It sounded just the same as it would have had I amplified it as usual.

BUT THEN. The next time I recorded, I noticed as soon as I stopped recording that the waveform looked super quiet. And I did my normal 8x amplification, and it was still way too quiet. So I had to multiply it again to get it to the right volume, and of course it sounded [bad] (having been amplified to such a degree) and I had to just trash the whole track. Why did it randomly record so quietly? (Coincidentally after having randomly recorded so loudly the previous recording.) How can I make sure this never happens again?

Your post was edited to be more family friendly.

How can I make sure this never happens again?

Audacity doesn’t apply effects or corrections when it records. What other audio applications do you use? Skype is famous for changing your computer sound settings “in the background” without telling you. Sometimes, they get put back to normal and sometimes not.

Do Not leave any other sound applications running while you try to record. Games can do this, too.

If needed, restart your machine and make sure nothing else starts. Use clean Shift-Shutdown and not restart. It takes longer, but more things get reset.

If you’re traditionally plagued with low volume, some Windows versions have a Microphone Boost setting in sound control panels.


The Snowball has a -10dB pad. You shouldn’t use that switch-position unless you are recording extra-loud sounds.

Also, check [u]Windows Microphone Boost[/u].

when I amplified the track to my normal multiplier (8), it clipped everywhere. So I undid it and realized that it had magically recorded my narration at an 8x amplification.

I assume you mean 8dB? (FYI… +8dB is an amplitude gain of about 2.5X)

Anyway… It’s usually best to adjust to a target level rather an by a fixed amount. Audacity will pre-scan your file and the Amplify effect will default to whatever gain is needed for normalized (AKA “maximized”) 0dB peaks.

The peaks don’t correlate well with perceived loudness, but often normalizing is what you want, or it can be a “good start”.

There are plug-ins for measuring RMS or LUFS, which better-represent “loudness” but you still have to be careful not to clip the peaks. And, if you are doing any processing/effects/editing that can affect loudness, it’s a good idea to run Amplify (or Normalize) before exporting to make sure you’re not clipping.

I meant multiply it by 8–as in I highlight the whole track, go to Effects, and Amplify, Allow Clipping, then type 8 in the field. Then I just delete or reduce the volume on any parts that got clipped. Anyway, it’s not important.

The good news is I think I solved my issue. The problem appears to have been that there is a switch on the back of the Blue Snowball, which I had thought had to do with what direction you wanted to record (one person, two people sat opposite, or a whole room). But as I slid it to each of the three positions as I was recording and looking at the waveform, the change in recorded volume was stark. So I must have usually kept it in the 1 position (medium volume), and the last couple times I had it between the 2 and 3 position, so that the first session it was more toward the 3 (loud), and the next time it had shifted toward the 2 position (quiet). I don’t know what the difference is between those in either decibels or “multipliers of amplification.” It seems like just a coincidence that recording in the 3 position is the same volume (maybe a touch louder) as recording in the 1 position and amplifying by 8x.

Anyway thanks for taking the time to try to help.

That’s not supported by the Amplify effect, which uses standard “dB” gain factors, but you can do that in the Nyquist Prompt with:

(mult *track* 8)

(fractional and negative values also work).

The omnidirectional setting of the Yeti (2) has about a 10dB volume boost compared to the other three settings. Sound doubles every 6dB.

I blame the manual. They never say that in English words. You have to interpret the associated engineering graph.

Conspiracy Theory says it’s an engineering mistake they’re trying to bury.