I have a long training course, so put together some pleasant backing music, mainly chimes and hums produced with Jukedeck’s automatic algorithm before they got sold. I don’t want it to sound repetitive, so put together 30 minutes of it, which I will then reuse.
When I put it together with the audio for the course, in Camtasia, the loudness is just too variable. For example, if in Camtasia I put the gain on the music at 20% to make it background, for some sections it is perfect, other stretches barely audible. However, if I put the gain at 30%, while some sections are nice background, other sections are overpoweringly loud.
I assume the Compressor is the answer, an effect I have not yet used, and don’t really understand. After normalizing the track, I have tried the compressor with the default settings:
Threshold: -12 dB
Noise Floor: -40 dB
Ratio: 2:1
Attack Time: 0.2 s
Release time: 1.0 s
I have tried applying it several times in a row. However, there are still sections that sound significantly louder than others when I try to make it background in Camtasia. Any suggestions on how to ensure it has a very very even level loudness across the entire track?
A compressor is the answer,
preferably one which takes into account the “psychoacoustic” peculiarities of human hearing.
A real-time compressor plug-in is better than Audacity’s native one.
as you can adjust the parameters as it’s playing, e.g.
If the threshold is too high the compressor won’t have any effect, (no matter how many times it is applied).
If you want the music to have next-to-no variation in volume, increase the compression ratio to 10:1.
Thanks so much. Very interesting link on the psychoacoustic peculiarities of human hearing.
I’d like to keep this as simple as possible to remain within the boundaries of my cognitive abilities Therefore, don’t think I’m capable of operating a plugin with adjustment as playing. Since this is background, I also have wide quality latitude and don’t need flawless results. I just need the same volume across the whole 30 minutes. So I’d prefer to use the builtin plugin if it will get me those results.
Can you help me understand a bit more your two comments, about threshold and compression ratio. Are these linked or separate?
That is, should I set “Ratio” to 10:1 instead of 2:1?
And should I also make an adjustment to the “Threshold” setting, and if yes, how should I decide what to use?
Chris wrote his custom compressor so he could listen to opera in the car. It’s job is to even out volumes with look-ahead tools so it doesn’t appear to be doing anything, but the show is magically even. Everything between full orchestra and one tenor in the south forty comes out about the same volume.
I used it for a long time to process a radio show I really liked. While it was broadcast, it was fine because it went through the broadcast compressors. When when they went on-line it was a disaster. One performer mumbled in his beer and the other had a thermo-nuclear laugh. No compression. I apply Chris’s Compressor with the first value, compression ratio, upped from 0.5 to 0.77. it produces a very high quality radio show without the radio noise.
compress.ny (16.9 KB)
Chris is still super handy because it doesn’t matter the show’s starting volume. Just get close and Chris will take it from there.
It evens everything out. Some normal Audacity compression tools have to be tailored to work right.
The only known shortcoming is the need to add “something,” some sound, to both ends of the show so Chris has something to chew on. It hates running off the ends of a file. Cut off the extra when you’re done. This is unlikely to get fixed because Chris reached End Of Life. So we just deal with it.
Very interesting, I’ll give it a try. When you say you need to add something to the end, I don’t need every last second, and can truncate some no problem, so am thinking of just running it on the track as is, and then cutting off some from the end. That work? If so, how much will I need to cut from the end?
Actually, I have three questions, hope you don’t mind:
Re truncation as per previous message, can I just truncate the last few seconds?
The volume differences in this background music are across long sections. That is, within any few seconds the volume is even. The issue is that one or two minutes are a bit quieter, then another stretch of one or two minutes are a bit louder. Does the “look ahead” operate across the whole thing, so the whole 30 odd minutes will be brought to about the same volume?
As per your advice, I can up the first value, compression ratio, from 0.5 to 0.77 and leave the rest as is. However, I am wondering if these settings are dependent on the normalization applied beforehand. Is there some normalization I should apply first, and if so to what dB?
it’s usually best to normalize to 0dB before applying the compressor.
And that’s the step that Chris doesn’t need. There is no requirement that the show get within range first. It’s not open-ended, however. You can’t put trash in there and it won’t solve overload.
When I was doing the show, I always got too much stuff at both ends of the download automatically, and so my extras on the ends were built-in. I know what happens at the end of the show. Chris sees the silence after the file is finished (look-ahead, right?) and goes nuts for the last few seconds. You want the nuts-going to happen in your trash segment not the show.
There is one other possible problem. I’ve never used it on shows longer than an hour. Some Audacity tools don’t like super long shows very much. That leads to the admonition that Audacity doesn’t make a very good surveillance recorder.
Ok. Given the main issue - that it is pretty even within any given section several seconds long, and the main problem is between differences many seconds apart - would this suggest normalization would still be a good first step?
In other words, if it just evens the volume within any given few seconds, but minute 1 is still much quieter than louder minute 5, which is much louder than minute 7, etc., it won’t solve my problem. I need an even volume across the whole 30 minute span. I don’t care how loud that is, as long as it is the same for the entire 30 minutes.
Normalization, (RMS or Peak), applies a constant gain-change cross the entire track:
Normalization does not adapt during the track to the changing volume of the track.
Compression does adapt, as does AGC & Level speech.
I guess my key question is still - do any of these plugins even out the loudness of sections several minutes apart?
This is probably poorly described, however something like: all volume for the entire track shall be within the following narrow range. Or, all volume of the track shall be adjusted to be within a small delta of the peak.
If one listens to any given ten second section, it sounds ok. But the volume is slowly changing across the track, sometimes louder, sometimes softer, so some sections eventually sound louder than others, after a minute or two. But this difference can only be heard in sections a minute or more apart.
When I put that track behind the spoken audio track though, the difficulty arises. No matter what I set the gain to, either some sections of the music are too quiet, or too loud.
Compression, AGC, & LevelSpeech will all reduce the dynamic-range.
Usually downward compression: sounds above a threshold are turned down.
Another option is to AutoDuck the music when the person speaks, so they can be understood.
(IMO Steve’s Dynamic Mirror plugin is easier to use than Audacity’s Native AutoDuck).
I tried Chris’s Compressor, applied up to five times, however am still finding significant differences in volume in sections several minutes apart. So my inference is that it evens out the volume within any given section of a few seconds, but not across minutes at a time. Please correct if I have this wrong.
However, I’m not sure how to set the threshold based on RMS. I note “RMS” was just written on the pic in red
I am using Audacity 2.2.2 on Windows for reasons of compatibility with existing work. Is there another version or plugin I should use to use the RMS approach you describe?
Ok, I think I’ve figured it out. By default it is using RMS compression. It is only if you select “Compressed based on peaks” that it does not. I will experiment a bit.
There is a tool in Audacity to measure the RMS volume of selected audio … Measure RMS - Audacity Manual
(RMS is just a suggestion:
you may need to use a more negative value than RMS for the compressor threshold).
Thanks. That plugin is not listed in the plug-in manager. I also looked under “Measure RMS” and it is not there either. Is there somewhere I can get it?
Therefore, don’t think I’m capable of operating a plugin with adjustment as playing. Since this is background, I also have wide quality latitude and don’t need flawless results. I just need the same volume across the whole 30 minutes. So I’d prefer to use the builtin plugin if it will get me those results.
