I have an audio which has loud parts with speech (which are very close to 0dB) and it has quiet parts with speech (which are very close to -40dB). I need to increase the volume only of the sounds below -20dB.
I tried using a compressor to make the loud parts quieter and then I normalized everything, but it messed the sound of the loud parts. Without using compressor, the sound of the loud parts is perfectly clear, very nice and looks like “high definition”. When I use the compressor the sound of the loud parts never gets as good as it was originally (I tried many different settings and presets). So, the best approach here is to increase the volume of the quiet parts, without changing anything in the loud parts.
GoldWave’s compressor/expander has a “boost quiet parts” preset. You can adjust the threshold and ratio (what they call “multiplier”) after selecting the preset.
GoldWave isn’t free but there is a fully-functional free trial and hopefully this isn’t something you need to do all the time!
I wouldn’t expect great quality since you’re boosting any background noise. And it won’t work if there are other sounds “masking” the quiet sounds.
Note that GoldWave sometimes mixes-up the definition of “compression” and “expansion” but the presets do what they say. This is actually upward compression… You are compressing/reducing the dynamic range by bringing the loud & quiet parts closer together.
Thank you so much guys! I hoped this would be possible without plugins (if it’s not possible, I will give it a try with the 2 plugins you recommended). I was thinking about using the “Analyze” feature of Audacity and apply effects to the analyzed parts which are below a threshold (then I would create a macro later), but even still it’s not very easy.
Anyway, if you are aware of some effect already present in audcatiy that helps me with this, I am all ears.
Sometimes this works out better if we know what the final presentation is expected to be. What’s the show?
There’s a note that Audacity can’t be used for Conflict Resolution, Surveillance, or Law Enforcement. Among a lot of problems, it doesn’t have time of day tags or other information.
My client recorded a Senator here in Brazil saying some suspicious things. The voice of some people in the audio are very loud but the senator’s voice if very low. I dont want to change a “single bit” from the voices of people that are loud enought because even a small change in it could be seen as “edited audio” which would be discarded as proof. However, the voice from the senator I need to increase the volume and in this case, every single time the senator spearks it will be written in the judicial records that the audio from that exactly parts had the volume increase without any other change.
Using compressor is unaceptable in this case because it will mess with the entire audio, it will change lots of things in “unexpected ways” (I know the compressor algorithm is deterministic however it changes so much the audio that anyone could say the audio was heavily “edited”). That’s why I would like to just increase the volume of the parts really in need and dont touch anything else in the recording.
High lite the low speech and normalize then compress using heavy ratio and attack
times as well as threshold This will help blend the volumes and glue it together
If your good with the normalize then a lite comp should be good.
If the speech over lap then compression is about all you have to work with
You need to use multiple compression ( little at a time ) to keep pumping down
Which means different settings each use…
Another option is cut and splice or reprint
By placing one vocal on left channel and other on the right
you will have to make two channels and pan them as so print it
bring up the new file(s) and start selecting and lowering opposing vocals
so now each vocal is left and right or dual mono which is most likely to happen
with the lower vocals go back and raise them and reprint your file again
Now with two separate files
I would mono this and compress a little Or use a stereo tool to blend the both sides
to your liking
You are now a master editor … LOL
ALSO,
Possible some Eq to help control back ground noise as well as pumping also
Compressing 40 db is a lot so noise and distortion is bound to happen
You could try hard limit then compress then repeat
Use the compression to bring the volume down
This is still kind of aggressive …
Good Luck
Mike
If you just select and amplify or normalize the quiet parts, that’s probably the best you can do. And the volume boost will be limited by the loudest part of the selection. If the background noise dominates, it will still dominate. If it’s still unintelligible after boosting, any processing beyond that is unlikely to help. The human brain is often the best “filter”.
It’s probably going to be obvious that you messed with it (because the background noise will go up-and-down) and they’ll probably accuse you of “editing” in any case, even if you don’t touch it!
If I understand, you want to reduce the loud 0db to -20db and quiet -40db closer to -20db.
You could try using a curve (move high and low range closer to -20), in a waveshaper, in a dynamic range compressor, or changing the level of tone (whatever this effect is called, brightness in a grayscale spectrum is changed in gamma).
You might be able to find plugins to do this.
Or you can do it manually with an envolope and fade the volume overtime to the desired level.
Quote: Because reverse is slower than forward; it is speed in negative.