Hi.
As some will know I'm busy with that collaboration project over on the Seymour Duncan forums.
In short: different guitarists are sending to me their solos that they've worked on and recorded on their own gear and it's my function to put the whole lot together nice.
But I'm running into a problem:
Obviously each guitarist's gear etc. is different as is their guitar tone and recording methods (levels to be exact). So when I pull all of the solos onto the backing track the solos are not at the same PERCEIVED VOLUME. I'm emphasizing the words "PERCEIVED VOLUME" because this is NOT (or so it would seem anyway) the same thing as the LEVELS of the audio. In other words: I can use any number of plugins that will even out the LEVELS (peak and RMS) but it does NOT follow that even with the LEVELS having being evened out that the PERCEIVED VOLUME is the same. Put another way: one guitarist may have used a tone with less tops and more bottoms while another's tone may have more treble and less bottom end. The one with the more treble is PERCEIVED to be louder in the mix because of the higher frequencies.
So first off: anybody got any suggestions as to how they'd adjust the individual solos so that they're PERCEIVED to be at the same volume???
I can tell you that something like HornetVUMeter does a terrible job here as it appears as though it will set the level of the track to EITHER -6dB peak OR -18dB RMS. Problem here is that in some instances: the RMS of the original audio will be less than -18dB RMS but the peaks will be greater than -6dB and in this case HornetVUMeter will reduce the peaks to -6dB but this results also in reducing the RMS level even more than the original and of course this audio is then softer than say another solo or guitar part.
The closest I've come to evening out these things has been in Sound Forge Pro 12. Basically I normalize the RMS levels to -18dB for each solo and then use something called "Wave Hammer" to limit the peaks to -6dB. But it concerns me that this may not be the way to do it because for sure I'm messing about with the dynamic range of the original audio. Furthermore: surely I don't have to resort to Sound Forge Pro 12 to be able to do this i.e. surely there must be a plugin or something that can achieve the same thing in the mix??? Or not???
Now one thing I've done some reading up on in this metering system of "LUFS" which is supposed to be PERCEIVED loudness. But my question is: on what is that based??? Does that factor into the equation that human hearing is more susceptible to certain frequencies than others???
Also wondering if surely there is a plugin that limits as well as compresses??? The idea being that one would compress the audio so that the resultant LUFS is at a certain level (say -12 LUFS) while at the same time limiting the peaks (here also an issue i.e. "peaks" or "true peaks"???) to -dB. Chances are that with the gazillion plugins that I have: I do have something like this. Just don't know where to look (I got so many compressors and limiters it's just not funny anymore).
Thoughts??? Input???
Consistent or uniform levels for audio clips
Moderators: Acoustica Greg, Acoustica Eric, Acoustica Dan, rsaintjohn
Re: Consistent or uniform levels for audio clips
Hmmmnnn...
As usual I may have answered my own question:
https://www.youtube.com/watch?v=-iKJF1xAQtY
Still. Thoughts and input would be appreciated.
As usual I may have answered my own question:
https://www.youtube.com/watch?v=-iKJF1xAQtY
Still. Thoughts and input would be appreciated.
Re: Consistent or uniform levels for audio clips
Who said mixing was easy again??? LOL!!!
Dunno why I ask these questions because no sooner do I ask them and then I find the solution!!! LOL!!!
If anybody here thinks that proper gain staging and level control is as simply as slapping something like Hornet's HornetVUMeter on each track and telling it to "learn and adjust" then think again!!! All it does is adjust the tack level to hammer either the peaks down to -6dB (and takes RMS and LUFS with it) OR it hammers RMS to -18dB and takes the peaks down with it. Either way: it's not matching track levels.
The solution is (drum roll):
EITHER use something like Hornet's HornetTheNormalizer (knew there was a reason for my buying this plugin when I had money floating around) OR (and this is the better option) use Sonalkasis FreeG and iZotope Insight and manually adjust the LUFS for each track. What this does is NOT interfere with the dynamic range of the track but the PERCEIVED loudness of the tracks become the same. In my project: it just so happens that by adjusting the LUFS to -12dB for each track the peaks are as near as damnit to -6dB at the same time. There is one that was slightly over so just used Sound Forge Pro 12 to hammer the peaks of that track down to -6dB (but so slight I probably need not have bothered).
Hornet's HornetTheNormalizer does not work as well as doing it manually though. From what I can see: it's drawback is that it only measures in increments of 1dB. So if it detects one track as being at -11.9 LUFS it will boost it by +1dB which then results in the level being set at -10.9 LUFS (which is an entire 0.9 LUFS too much boost). At best it's an approximation.
The video gave me a good clue though. When setting these levels: for each track loop a section of the track and let iZotope average the INTERNAL LUFS (I ran it for about two minutes or so on short portions of each clip and waited for it to settle down with the right value). If needs be then just adjust the track level with the Snalkasis FreeG plugin and re-test.
One thing to take note of: if you're dealing with guitar tracks it's important to bypass the initial attach portions of each take though (for this reason it may be a worthwhile idea to apply come compression to the tracks to eliminate these attack peaks but for me this is again interfering with the dynamic range).
OBVIOUSLY once you've got the correct track levels set (according to metering) well then it's time to use your ears. But so far as I can tell: the above puts you pretty much right in the correct ballpark.
...
Dunno why I ask these questions because no sooner do I ask them and then I find the solution!!! LOL!!!
If anybody here thinks that proper gain staging and level control is as simply as slapping something like Hornet's HornetVUMeter on each track and telling it to "learn and adjust" then think again!!! All it does is adjust the tack level to hammer either the peaks down to -6dB (and takes RMS and LUFS with it) OR it hammers RMS to -18dB and takes the peaks down with it. Either way: it's not matching track levels.
The solution is (drum roll):
EITHER use something like Hornet's HornetTheNormalizer (knew there was a reason for my buying this plugin when I had money floating around) OR (and this is the better option) use Sonalkasis FreeG and iZotope Insight and manually adjust the LUFS for each track. What this does is NOT interfere with the dynamic range of the track but the PERCEIVED loudness of the tracks become the same. In my project: it just so happens that by adjusting the LUFS to -12dB for each track the peaks are as near as damnit to -6dB at the same time. There is one that was slightly over so just used Sound Forge Pro 12 to hammer the peaks of that track down to -6dB (but so slight I probably need not have bothered).
Hornet's HornetTheNormalizer does not work as well as doing it manually though. From what I can see: it's drawback is that it only measures in increments of 1dB. So if it detects one track as being at -11.9 LUFS it will boost it by +1dB which then results in the level being set at -10.9 LUFS (which is an entire 0.9 LUFS too much boost). At best it's an approximation.
The video gave me a good clue though. When setting these levels: for each track loop a section of the track and let iZotope average the INTERNAL LUFS (I ran it for about two minutes or so on short portions of each clip and waited for it to settle down with the right value). If needs be then just adjust the track level with the Snalkasis FreeG plugin and re-test.
One thing to take note of: if you're dealing with guitar tracks it's important to bypass the initial attach portions of each take though (for this reason it may be a worthwhile idea to apply come compression to the tracks to eliminate these attack peaks but for me this is again interfering with the dynamic range).
OBVIOUSLY once you've got the correct track levels set (according to metering) well then it's time to use your ears. But so far as I can tell: the above puts you pretty much right in the correct ballpark.
...
-
Mark Bliss
- Posts: 7313
- Joined: Fri Jan 20, 2012 3:59 pm
- Location: Out there
Re: Consistent or uniform levels for audio clips
How you feeling Dale? I sense a little tension...... 
Methods vary. Situations vary. There is no "blanket correct answer"
First of all, going back to your first post: Volume vs loudness confusion and clarification for any who might care to follow along.
Volume = turning up the knob on an amplifier. Thank goodness for the volume knob.
"Perceived loudness" is a whole 'nuther thing. (Or a hole in which one unfortunately falls.
)
Loudness has little or nothing to do with peaks and is all about what I prefer to refer to as the "density" of the audio signal. "Back in the day" we had little to work with beyond funky little Vu meters and a rough general idea of where the ideal level range was. But it worked well if one payed attention.
RMS is a method of measuring the audio signal, but it was really more of a lab reference or equipment spec. Most didnt have a method to measure it. (And for if interested, it is a mathmatic formula of the average voltage of an audio signal. "Root Mean Squared") Really only in the modern era of digital gear did the consumer commonly have access to measuring this.
And as pointed out, as a measurement it does not compensate for the perception of loudness across the frequency spectrum (RE: "Fletcher-Munson curve") Only very recently has the Lu measurement been introduced as an effort to compensate.
Second post and in general: Surprise! I think your over thinking this.
The recording stage is done, so we skip that concern of gain staging.
Yes, it is considered good practice to go into the inserts at about -18dB But its not critical to the accuracy of 1dB IMO, just a general range. Peaks shouldn't be a big concern here at all.
And in this particular situation, the processing is largely done right? So perhaps its even a bit less critical.
I am not sure why Saverio chose to set up his Hornet plug in the way he did. I've used it some, thought it worked well and sped up the process for my purposes on raw unprocessed (or minimally processed) recorded tracks when setting up a mix, but I don't think its the desired tool for this intent.
And I cant recall at the moment, I don't think you can turn off the -6 peak component, which harms the plug ins utility in cases where you want to ignore peaks. There are other solutions. Such as you mention- a tool that normalizes to a user defined RMS or Lu, (Not peak) level, if that was desired.
But in this case, I don't think you need any level or gain plug in at all.....
In this case, I see no point in -6 peaks at the track level. That's really just a master bus headroom target for me. And to hit that, the tracks would likely have to be attenuated at some stage of the process anyway.
Also, if the tracks are that dynamic, perhaps using an average track loudness level isn't ideal anyway. I often would just take a quick look at the loudest portion of a track and make sure I am in a general ballpark target range.
Bottom line is that in this situation, you are mixing the output of the tracks right? So what matters is what you hear, not some technical level setting. Surely "all equal perceived loudness" is not the goal. Wouldn't lead to a very interesting mix IMO. Just mix the tracks. The recording levels are done and gone. The desired tone/timbre is done, and you don't want to do anything major that would color that.
And the cumulative level is probably going to sum together and exceed the desired target range on the master bus, so that's where I would be focusing my attention on levels, whatever scale you are looking at.
Yes, you get there largely from track levels and track loudness, not master bus processing, but.... Fixed or specified track target levels has never once worked for me.
Methods vary. Situations vary. There is no "blanket correct answer"
First of all, going back to your first post: Volume vs loudness confusion and clarification for any who might care to follow along.
Volume = turning up the knob on an amplifier. Thank goodness for the volume knob.
"Perceived loudness" is a whole 'nuther thing. (Or a hole in which one unfortunately falls.
)Loudness has little or nothing to do with peaks and is all about what I prefer to refer to as the "density" of the audio signal. "Back in the day" we had little to work with beyond funky little Vu meters and a rough general idea of where the ideal level range was. But it worked well if one payed attention.
RMS is a method of measuring the audio signal, but it was really more of a lab reference or equipment spec. Most didnt have a method to measure it. (And for if interested, it is a mathmatic formula of the average voltage of an audio signal. "Root Mean Squared") Really only in the modern era of digital gear did the consumer commonly have access to measuring this.
And as pointed out, as a measurement it does not compensate for the perception of loudness across the frequency spectrum (RE: "Fletcher-Munson curve") Only very recently has the Lu measurement been introduced as an effort to compensate.
Second post and in general: Surprise! I think your over thinking this.
The recording stage is done, so we skip that concern of gain staging.
Yes, it is considered good practice to go into the inserts at about -18dB But its not critical to the accuracy of 1dB IMO, just a general range. Peaks shouldn't be a big concern here at all.
And in this particular situation, the processing is largely done right? So perhaps its even a bit less critical.
I am not sure why Saverio chose to set up his Hornet plug in the way he did. I've used it some, thought it worked well and sped up the process for my purposes on raw unprocessed (or minimally processed) recorded tracks when setting up a mix, but I don't think its the desired tool for this intent.
And I cant recall at the moment, I don't think you can turn off the -6 peak component, which harms the plug ins utility in cases where you want to ignore peaks. There are other solutions. Such as you mention- a tool that normalizes to a user defined RMS or Lu, (Not peak) level, if that was desired.
But in this case, I don't think you need any level or gain plug in at all.....
In this case, I see no point in -6 peaks at the track level. That's really just a master bus headroom target for me. And to hit that, the tracks would likely have to be attenuated at some stage of the process anyway.
Also, if the tracks are that dynamic, perhaps using an average track loudness level isn't ideal anyway. I often would just take a quick look at the loudest portion of a track and make sure I am in a general ballpark target range.
Bottom line is that in this situation, you are mixing the output of the tracks right? So what matters is what you hear, not some technical level setting. Surely "all equal perceived loudness" is not the goal. Wouldn't lead to a very interesting mix IMO. Just mix the tracks. The recording levels are done and gone. The desired tone/timbre is done, and you don't want to do anything major that would color that.
And the cumulative level is probably going to sum together and exceed the desired target range on the master bus, so that's where I would be focusing my attention on levels, whatever scale you are looking at.
Yes, you get there largely from track levels and track loudness, not master bus processing, but.... Fixed or specified track target levels has never once worked for me.
Re: Consistent or uniform levels for audio clips
FWIW: I use left/right on a scope and my ear to compare guitar volumes/frequency/tones and manually adjust them.
- Joe -
MX9PS, 64-bit, build 460. Windows 10, Intel i5, 64-bit, 1.8GHz, 8GB.
Reverbnation: https://www.reverbnation.com/joelouvar
MX9PS, 64-bit, build 460. Windows 10, Intel i5, 64-bit, 1.8GHz, 8GB.
Reverbnation: https://www.reverbnation.com/joelouvar
-
Mark Bliss
- Posts: 7313
- Joined: Fri Jan 20, 2012 3:59 pm
- Location: Out there
Re: Consistent or uniform levels for audio clips
@Dale:
Hmm. Backing up and re-reading, perhaps I misunderstood the discerning point between "tracks" and "clips"
It doesn't change much of the reply however. I'd put them in their own tracks.
But you could use clips to assemble a track, and adjust them relatively, starting with clip gain.
Maybe I am missing something.
Hmm. Backing up and re-reading, perhaps I misunderstood the discerning point between "tracks" and "clips"
It doesn't change much of the reply however. I'd put them in their own tracks.
But you could use clips to assemble a track, and adjust them relatively, starting with clip gain.
Maybe I am missing something.
Re: Consistent or uniform levels for audio clips
Hello.
Nope. You got it right really. Each guitar solo is on its own track. But it gets a bit more involved because one guitarist may have done multiple solos in the same take and on the same track but with different amp. settings and therefore a different tone and even different levels on the same track (so in this case I've had to split the track and move each solo to its own track which is fine).
Oh and very few of these dudes are adhering to any particular level by the looks of things which also makes it more difficult. For one thing: the main guitar track which at the moment includes the backing track is extremely hot (you can actually hear some distortion if you listen carefully). So I attenuate that of course (but the distortion is already built in) (but see below the reason for this track at the moment).
It would obviously be much easier if all were in the same studio and I was doing the recording (as the original was done) but alas: no such luxury for this.
Been doing a boatload of experimentation and the best I can come up with seems to be using iZotope's Insight with that Sonalkasis FreeG plugin. For each audio clip (which may or may not be an entire track as noted above) I'm just ensuring that the each has pretty much the same LUFS readings. Once that's all done then I'm sort of "adjusting to taste" really. Problem is you have to get in the ballpark first though i.e. to try do this with this many different audio clips from different parts of the world isn't easy i.e. you get one or two that sound right and the the others go out of whack.
The final levels are pretty important too though i.e. as you may remember from my testing for mastering purposes I've found that what works best is to send the final mix at -6dB peak/-18db RMS to the mastering software and then push it in there if needs be.
You can disable that stuff in HornetVUMeter by the way e.g. you can set peaks to 0dB while leaving RMS at -18dB (I know you can do this but I never actually tried this today).
Of particular interest to me though: there's much to this LUFS thingy. For the purposes of the track I'm panning the different guitar solos slightly left and slightly right as they come. iZotope's Insight actually picks this up. In other words: a track where both sides are of equal PERCEIVED loudness it may give a reading of X. But the moment you pan the track that reading of X changes (actually drops). That's quite interesting of course and makes total sense. Not bad stuff this iZotope stuff either!!! LOL!!!
I'd tried a "cheat method" too i.e. put a compressor on each track and simply force the output to be equal for the whole bunch. But I got the result that you pointed out i.e. everything came to the exact same level so the end result was that the solos got lost in the final track i.e. everything came out on the same level.
Anyway. I'm getting there.
For the sake of interest here's (below) a link to the post where I've just uploaded this stuff. You tell me what you think e.g. does it sound like a coherent type of mix or what???
https://www.seymourduncan.com/forum/sho ... ost4310084
Sorry. The song is LONG and the solos only start about half way through but bear with me!!! LOL!!!
There are some issues. For one thing I'm still working with .MP3 files until everybody has done their thing (in .WAV format this thing is 170MB per track). Also: the backing track I'm using is not the original MIDI track with VST drums and bass i.e. it's the very first track that was laid down by the first guitarist so at this stage and in order to test the doubled guitar rhythm track the bass and drums are also unfortunately being panned with the first main guitar track (but this will sort itself out when I get all of the individual guitar tracks in .WAV format WITHOUT the backing track) (hope that makes sense).
Nope. You got it right really. Each guitar solo is on its own track. But it gets a bit more involved because one guitarist may have done multiple solos in the same take and on the same track but with different amp. settings and therefore a different tone and even different levels on the same track (so in this case I've had to split the track and move each solo to its own track which is fine).
Oh and very few of these dudes are adhering to any particular level by the looks of things which also makes it more difficult. For one thing: the main guitar track which at the moment includes the backing track is extremely hot (you can actually hear some distortion if you listen carefully). So I attenuate that of course (but the distortion is already built in) (but see below the reason for this track at the moment).
It would obviously be much easier if all were in the same studio and I was doing the recording (as the original was done) but alas: no such luxury for this.
Been doing a boatload of experimentation and the best I can come up with seems to be using iZotope's Insight with that Sonalkasis FreeG plugin. For each audio clip (which may or may not be an entire track as noted above) I'm just ensuring that the each has pretty much the same LUFS readings. Once that's all done then I'm sort of "adjusting to taste" really. Problem is you have to get in the ballpark first though i.e. to try do this with this many different audio clips from different parts of the world isn't easy i.e. you get one or two that sound right and the the others go out of whack.
The final levels are pretty important too though i.e. as you may remember from my testing for mastering purposes I've found that what works best is to send the final mix at -6dB peak/-18db RMS to the mastering software and then push it in there if needs be.
You can disable that stuff in HornetVUMeter by the way e.g. you can set peaks to 0dB while leaving RMS at -18dB (I know you can do this but I never actually tried this today).
Of particular interest to me though: there's much to this LUFS thingy. For the purposes of the track I'm panning the different guitar solos slightly left and slightly right as they come. iZotope's Insight actually picks this up. In other words: a track where both sides are of equal PERCEIVED loudness it may give a reading of X. But the moment you pan the track that reading of X changes (actually drops). That's quite interesting of course and makes total sense. Not bad stuff this iZotope stuff either!!! LOL!!!
I'd tried a "cheat method" too i.e. put a compressor on each track and simply force the output to be equal for the whole bunch. But I got the result that you pointed out i.e. everything came to the exact same level so the end result was that the solos got lost in the final track i.e. everything came out on the same level.
Anyway. I'm getting there.
For the sake of interest here's (below) a link to the post where I've just uploaded this stuff. You tell me what you think e.g. does it sound like a coherent type of mix or what???
https://www.seymourduncan.com/forum/sho ... ost4310084
Sorry. The song is LONG and the solos only start about half way through but bear with me!!! LOL!!!
There are some issues. For one thing I'm still working with .MP3 files until everybody has done their thing (in .WAV format this thing is 170MB per track). Also: the backing track I'm using is not the original MIDI track with VST drums and bass i.e. it's the very first track that was laid down by the first guitarist so at this stage and in order to test the doubled guitar rhythm track the bass and drums are also unfortunately being panned with the first main guitar track (but this will sort itself out when I get all of the individual guitar tracks in .WAV format WITHOUT the backing track) (hope that makes sense).
Re: Consistent or uniform levels for audio clips
Morning.
As a matter of interest: is there any way to place multiple instances of a plugin on a track this as opposed to one plugin on a track that applies to the entire track??? And each instance will apply only to a specified portion of the track??? Note this not the same as automation. With an EQ plugin for example: I think you can use automation to apply different EQ to different portions of the track using the same and one instance of the EQ plugin (am I right???). But I’m talking about plugins that have to analyze or “learn” and have to be initialized each time before beginning the analysis or “learning”.
As a matter of interest: is there any way to place multiple instances of a plugin on a track this as opposed to one plugin on a track that applies to the entire track??? And each instance will apply only to a specified portion of the track??? Note this not the same as automation. With an EQ plugin for example: I think you can use automation to apply different EQ to different portions of the track using the same and one instance of the EQ plugin (am I right???). But I’m talking about plugins that have to analyze or “learn” and have to be initialized each time before beginning the analysis or “learning”.
-
freightgod
- Posts: 720
- Joined: Thu Nov 29, 2012 5:18 am
Re: Consistent or uniform levels for audio clips
It's 3 AM! Get some sleep! No more coffee!dpaterson wrote:Morning.
As a matter of interest: is there any way to place multiple instances of a plugin on a track this as opposed to one plugin on a track that applies to the entire track??? And each instance will apply only to a specified portion of the track??? Note this not the same as automation. With an EQ plugin for example: I think you can use automation to apply different EQ to different portions of the track using the same and one instance of the EQ plugin (am I right???). But I’m talking about plugins that have to analyze or “learn” and have to be initialized each time before beginning the analysis or “learning”.
It was only at the end of your post when you talked about plugins that need something to chew on first that sense this made.
Re: Consistent or uniform levels for audio clips
By the time you posted that post I was already a whole day ahead of you!!! LOL!!!
Tell you something though: after some more playing around I’ve come to the conclusion that you cannot beat good ‘ol Sonic Foundry err... Sony err... MAGIX!!! LOL!!! I had been sent another two solos so thought hang on a minute: just make Sound Forge Pro 12 my external audio editor in Mixcraft. Piece of cake. Opened each track in Sound Forge and normalized the RMS levels to -24dB for all the bits and pieces i.e. either entire tracks or different pieces of audio on the same track. End result: something that borders on what I would call an almost perfectly balanced mix (at very least a magical starting point). Turns out that Sound Forge has a special little feature called “equal loudness contour” which so far as I can tell accomplishes the same thing as one would expect if using LUFS to set levels and it is as accurate as hell I tell you.
The problem with using plugins for this is that they adjust the levels based on the very first piece of audio that they analyze. So if a passage starts with an initial attack the plugin immediately sets the level based on that initial attack regardless of the fact that the rest of the audio for that particular piece of audio has a lower level than the initial attack.
Oh well. That’s as good as it gets. For today. Could change tomorrow!!! LOL!!!
Tell you something though: after some more playing around I’ve come to the conclusion that you cannot beat good ‘ol Sonic Foundry err... Sony err... MAGIX!!! LOL!!! I had been sent another two solos so thought hang on a minute: just make Sound Forge Pro 12 my external audio editor in Mixcraft. Piece of cake. Opened each track in Sound Forge and normalized the RMS levels to -24dB for all the bits and pieces i.e. either entire tracks or different pieces of audio on the same track. End result: something that borders on what I would call an almost perfectly balanced mix (at very least a magical starting point). Turns out that Sound Forge has a special little feature called “equal loudness contour” which so far as I can tell accomplishes the same thing as one would expect if using LUFS to set levels and it is as accurate as hell I tell you.
The problem with using plugins for this is that they adjust the levels based on the very first piece of audio that they analyze. So if a passage starts with an initial attack the plugin immediately sets the level based on that initial attack regardless of the fact that the rest of the audio for that particular piece of audio has a lower level than the initial attack.
Oh well. That’s as good as it gets. For today. Could change tomorrow!!! LOL!!!