Could use some workflow advice on how to best achieve this. I’m working with a media company that has specific specs on music delivery:
File Format: WAV Sample Rate: 48 kHz Bit Depth: 24-bit Integrated Loudness: Exactly -16 LKFS / LUFS. Absolute True Peak: Not louder than -2 dBTP (this can be less, i.e. - 6 dbTP). Track Top: Track must start after 1/10th of a second of silence (100 milliseconds). Track Tail: Track must have a natural tail and a fade to total silence (ie. no hiss). No gradual fade outs of continuous music.
Nothing too crazy there. I also have to output a bunch of Alt mixes, for example, underscore, no vocals etc. And also output stem breakouts. I already have the appropriate group busses in my template to make this straightforward enough, but I’m struggling with achieving a specific LUFS target in a streamlined way.
I’m trying to avoid 2 mix processing as I need the stems to accurately reflect the full mix. I can strap a peak limiter on the master to cut at -2DBTP as a safety for the full mix and/or stems. Obviously if I need anything on the master then I forfeit the option to export multiple outs in one go, which i don’t really want to do. I have about 20 cues and need anywhere from 3 to 6 alt mixes of each cue, plus stems. So I’m doing a lot of bouncing.
I can achieve something approximate to the target LUFS using series of group compressors/limiters but it’s just proving time-consuming, adjusting/re-adjusting , back and forth over and over until I get something near. Bouncing, checking Statistics, rinse and repeat.
How could I best go about this, without using any master bus processing? Obviously I can monitor integrated LUFS on the master but its difficult just using group processing. AFAIK, Cubase does not have a normalise to LUFS function, which might make life a bit easier.
Not a huge workflow input - but maybe you can try bouncing all tracks individually with normalising at -3db than your targetted LU (e.g. - normalise to -19 if tagetted is -16) and then render the final output and check the stats of the final render. (you can normalise at the level you want - but the normalise always takes previous value & normalises without waiting for you to change it - thats annoying, but once you normalise again with the setting you want - it starts normalising all with the target LU)
Cubase Pro does.
It is an option in the Normalize function, one of the Direct Offline Processes.
However, this works only on a rendered stereo track and in order to cope with this requirement…
…you’d need to apply a little trick.
Get the statistics for the unnormalized stereo mix file. Take note of the peak amplitude max. value.
Then normalize the file to -16 LUFS, open the stats again and compare the peak amplitude. You can now see by how many dB Cubase changed the file in order to get to the desired LUFS.
You can then go back to your full mix project and adjust the volume fader by the deducted amount.
Don’t forget that Steinberg’s Brickwall Limiter has intersample detection, so you can use that for the -2dB TP.
Ah yes, forgot to look in the Direct Offline processing.
Yeah I’m starting to realise there will be no magic bullet. Thanks I’ll try that next. I’m basically riding group levels to get a ballpark reading on the master bus (with Supervision), and fine tuning from there. It’s easier if i back off peak levels, and there’s no requirement that it has to be close to -2dbTP, just that it doesnt go over, so I can avoid master bus peak limiters easily enough.
But then I’m also getting into territory where the mixes are quite dynamic, around -21 or -20LUFS. And therefore require some compression and evening out of the dynamics to bring LUFS levels up, and to keep levels somewhat consistent. It’s all doable but just very time consuming.
I think the difficult thing was that these tracks were made and mixed well before I was submitting to these people and adhering to this spec, so i’ve gotta rejig them a bit more than usual from the get go.
I have stems already printed for most of them so can normalise the stems. I was wondering if there some magical mathematical formula to say, normalise 5 stems separately to a certain loudness level in order to achieve an overall loudness level when summed together.
The way you described the task is that your client asks you to master your project but without using mastering.
That stems need to reflect the final mix accurately is a weird requirement. No wonder you are spending a lot of time on this.
You can get close by calculating the result of summing stems using this calculator (fundamentally for SPL levels but should work for integrated LUFS values). It’s the other way round to what you’re asking for but it might help for some quick trial and error calculations.
Thanks, I couldn’t quite work out how to accommodate it for making a LUFS measurement. Sadly I don’t own Wavelab Pro currently (I use Acon Acoustica mainly).
After some experimentation it seems as though just trial and error, bouncing, tweaking, rebouncing, using group VCAs and a 2bus peak limiter is kinda the only way I can see.
You can simply enter an LUFS integrated measurement for each stem in the Level boxes (-20, -25, -14 or whatever) and then click on the ‘Calculate’ button. The result will give the overall integrated value when you mix these stems together. That’s all. It might help with trial and error but of course may not help at all - I’m not familiar with the precise workflow you are using. This assumes the values you enter are the integrated values including any channel fader adjustments. This calculator is certainly not some magical button to achieve what you require in one step.
As recommended above, don’t discount Goodhertz Loudness. This can help make the whole process easier. For true peak control I’d also recommend Steinberg’s Brickwall Limiter. For general transparent level control you might also find IK Multimedia’s Stealth Limiter helpful.
You might be right, at least where Cubase is concerned. The tips I posted were intended as help in that trial and error process. I too think there’s no magical tool which will do it for you except perhaps for Wavelab Pro’s Meta Normalizer (but if ever you intended to acquire this, you’d have to fully test beforehand to make sure it does what you require). Nuendo too might help with all this because of its loudness track.