I do think that a human engineer with experience can do better than landr. When I say mastering engineer I mean a pro who does that all day long and has experience and probably worked under the wing of someone else. That kind of person can get more out of the process for you.
Of course they can. An ML algorithm wouldn't have any point of reference if it weren't built on the work of the humans that perfected their craft long before it was developed.
Funny enough I see people frequently say that they believe
Neutron knows something they don't about the track(s) they apply it to. Neutron literally applies the same formula to everything you spit thorough it. It's a glorified preset algorithm with the ability to find a few resonances. (Even then it's not consistent).
It doesn't actually analyze or understand the crest factor of your music. It does't understand what an appropriate crest factor might be based on the genre or aesthetic you're going for. It doesn't understand if the exciter's an appropriate choice or not, and it doesn't know how much saturation would be appropriate to set.
Most importantly though, it's "choices" are repeatable once you know what the pattern to look for is. It's unbelievably easy to see just how
dumb it actually is if you put multiple instances in a row on the same track or bus.
It detects one of several instrument categories. The instrument categories are used to initiate the 1st layer of the preset. After it determines the instrument (or you define it), it then has
you define the final parameters of the preset by giving you 6 choices; 3 "Style" choices, 3 "Intensity" choices. The same instrument combined with the same style/intensity will always load the same preset configuration.
Warm and
Low will always spit out the same preset configuration,
Balanced and
High will always spit out the same preset configuration.
Etc.
Put 3 Neutrons in a row on the same channel and run the same exact configuration on each one. Using just the EQ as an example, while the nodes may move to different places (proving that it actually ins't "smart" at all), you'll notice the configuration is the same. If the configuration consists of a high-pass, a low shelf, two bell cuts, and a bell boost, you'll see the same configuration applied to each instance. Even the boost/cut amounts on all bands are identical, same with each band's Q width. Identical choices. It's preset box. None of its choices are informed.
Ozone works the same way. It will always set the same modules with the same general parameters. The numerical values may be different on each instance, but the configuration will be the same depending on the boxes you tick.
For example if you stacked several Ozones with the same settings in a row, and one of the results was a dynamic EQ with 4 bells and 1 shelf, all compressing; you'll see that all 3 instances make the same overall choice to add a dynamic EQ with 4 bells and 1 shelf all set to compress. If the dynamics module sets itself so only the low band compresses, all 3 instances will be set the same way. The way it defines which modules to set and how to set them up are based on the options you tick in the 'assistant' page.
The 'smartest' thing it does is aim for a target LUFS level.
This immediately raises the question:
If Ozone were actually attempting to achieve some kind of generic aesthetic standard/improvement then why does it stack the same processing 3 times in row instead of doing nothing, or close to nothing, with each successive instance inserted after the 1st one?
None of its choices are informed. I'd bet my retirement account that Landr (or any equivalent 'service') isn't any "smarter".
This is the technology people put their faith in as if it
knows something they don't.
3 INSTANCES OF NEUTRON 3 INSERTED ON THE SAME BUS:
ALL INSTANCES SET TO THE DEFAULT "Balanced" & "Medium".
Instance 1:
Instance 2:
Instance 3:
SO HOW DOES IT SOUND THEN?
Decide for yourself...
RAW DRUMS. (Slight processing, mixed to taste.)
Shared with Dropbox
www.dropbox.com
3 INSTANCES OF MIX ASSISTANT IN A ROW:
Shared with Dropbox
www.dropbox.com