Removing spikes with the compressor

Post your questions & get help from friendly LibriVoxers
Post Reply
Peter Why
Posts: 5815
Joined: November 24th, 2005, 3:54 am
Location: Chigwell (North-East London, U.K.)

Post by Peter Why »

Here's a strip of recording before and after compression: (I don't know how to get the image to display within a post - it's less than 9kb)

http://www.mediafire.com/file/a363m6y4c7bltnw/Comp_py.jpg

The compressor values for that second image are: threshold -16, Noise -40, Ratio 5:1. I've experimented with Ratio of 10:1, but it doesn't have any effect on that big spike.

Can anyone suggest how to get rid of spikes like that throughout a recording? At the moment I have to do it manually as I edit.

Peter
"I think, therefore I am, I think." Solomon Cohen, in Terry Pratchett's Dodger
TriciaG
LibriVox Admin Team
Posts: 60587
Joined: June 15th, 2008, 10:30 pm
Location: Toronto, ON (but Minnesotan to age 32)

Post by TriciaG »

I think Mediafire (and other file sharing sites) don't allow sharing of images on other sites - to cut down on their bandwidth. Here's your picture:

Image

It looks like you normalize after compression? I was expecting to see a smaller wave form.

Would you please upload a bit of the recording and post the link? I think experimentation would be easier than guessing. :)
Serial novel: The Wandering Jew
Medieval England meets Civil War Americans: Centuries Apart
Humor: My Lady Nicotine
TriciaG
LibriVox Admin Team
Posts: 60587
Joined: June 15th, 2008, 10:30 pm
Location: Toronto, ON (but Minnesotan to age 32)

Post by TriciaG »

Me again. :)

Just looking at the photo, the file DID compress the spike relative to the rest of the waveform. You cannot remove spikes altogether. :) However, if you want to do more, I suggest playing with the threshold. On our church sermon recordings (a very dynamic speaker - highs and lows), I use a threshold of -23 dB.
Serial novel: The Wandering Jew
Medieval England meets Civil War Americans: Centuries Apart
Humor: My Lady Nicotine
moniaqua
Posts: 1542
Joined: April 11th, 2013, 4:48 am
Location: Somewhere in the south

Post by moniaqua »

Peter Why wrote: February 15th, 2018, 4:41 am Can anyone suggest how to get rid of spikes like that throughout a recording?
Speaking less dynamically? - I know, it is not really an option :)

But if you know that there will be a louder word or sentence, you can back up from the microphone. Else - I have do de-amplify such spikes, too, if that is of any comfort for you.
tovarisch
Posts: 2936
Joined: February 24th, 2013, 7:14 am
Location: New Hampshire, USA

Post by tovarisch »

Monika,

Deamplifying a small section of the track around the spike is what I do, but that's what Peter wants to avoid - manual modification.

Peter,

How much do you see yourself spending on managing those spikes? I mean, if there were a perfect solution, how much time (percentage-wise) would you save versus addressing those loud pieces individually?
tovarisch
  • reality prompts me to scale down my reading, sorry to say
    to PLers: do correct my pronunciation please
moniaqua
Posts: 1542
Joined: April 11th, 2013, 4:48 am
Location: Somewhere in the south

Post by moniaqua »

tovarisch wrote: February 15th, 2018, 8:06 am Deamplifying a small section of the track around the spike is what I do, but that's what Peter wants to avoid - manual modification.
Yes, I thought so. I would love to avoid it, too, but the compressor does not work really good for me. Best really is to avoid big spikes, but sometimes that is difficult. Compressor often doesn't catch the big ones without altering the voice so that I don't like it (I apply it if necessary, though), but sometimes I still have to deal with spikes and modify them manually.
Peter Why
Posts: 5815
Joined: November 24th, 2005, 3:54 am
Location: Chigwell (North-East London, U.K.)

Post by Peter Why »

Thanks for your interest, all of you. Here's a short sample (seven seconds), unedited in any way.

https://librivox.org/uploads/tests/Mayhew_py.mp3

I'll go through your posts and add comments to this one shortly.

EDIT: Thanks again. You've got me experimenting with the compressor. I've just tried playing with the threshold, then with all the other parameters, without hitting an immediately obvious solution. But I found that if I compress with "compress based on peaks" checked (that's threshold -14, noise floor -40, ratio 5:1, next two set to minimum, and both check boxes ticked), THEN deamplify by 5dB, I get what looks and sounds like a very acceptable solution.

I'm not going to chance a whole recording without my prooflistener's approval of sound quality, but this may be the solution.

With regard to the amount of time I'm willing to spend, I do go through my whole recording, editing it manually, anyway, but being able to squeeze those spikes throughout my recording first will save me quite a bit of time: up til now I've put a lot of effort into selecting each spike and deamplifying it.

Peter
"I think, therefore I am, I think." Solomon Cohen, in Terry Pratchett's Dodger
Post Reply