I have a very important recording (impossible to rerecord now), which I screwed up making somehow. It’s as if the microphone constantly switched between two set distinct gain settings every second or so. So the result is like if you put a pulse wave envelope controlling the levels. (The attached image is a clip of continuous speech from one speaker; the flat bits are not just silence) I can hear the quieter parts with acceptable quality if I amplify by a huge amount, but the louder parts are fine as is so I don’t want to do anything to them.
Does audacity have a way to analyze the track and just select all of the quieter parts for amplification etc?
I’ve tried all sorts of combinations of amplification, compression and limiting, applied to the entire track, but the volume transitions are so extreme and fast that anything I do to the entire track ends up causing problems somewhere.
Does audacity have a way to analyze the track and just select all of the quieter parts for amplification etc?
Do you know that you’re going to get something good if you do that?
That looks like one of the Windows Enhancement tools trying to suppress background “noise.” If it suppresses far enough, it is noise and boosting it is just going to give you hissy, hummy, pumping words.
Can you take an example sentence and fix it manually? Carefully select the beginning and end of the quiet bits and apply effects and filters to make it match the normal interview voice? You will almost certainly need the zoom tools for this.
There are two pre-baked tools whose job it is to do stuff like this. LevelSpeech and Chris’s Compressor.
Chris developed his compressor so he could listen to opera in the car. It made him crazy when the show went from single violin down by the river to the whole orchestra blasting in his ear, so he developed a compressor to even things out. I increase the first number from 0.5 to 0.77 and it’s a perfect match to broadcast compressors.
The only oddity I know of is you need to add some extra show at the beginning and end and then cut it off later. Chris doesn’t like running off the ends.
Level Speech goes through and tries to level all the speeches—oddly enough.
I don’t know where the volume limits are for either tool.
It would be interesting to hear the result, if you could post a short clip containing some of the (originally) loud and (originally) quiet audio.
Around 20 seconds in MP3 or OGG format would be suitable for attaching to a post.
