Let's find out: Neutron beats human?

Now that computers can compose, arrange, play, record, mix and master music without human intervention, why does it even matter what human listeners think? If you want to know which mix is best, write some code to analyze them and make that decision.
 
With no sarcasm intended, did you listen that far?

I listened to all three...but TBH...I don't listen to one whole mix, then another...rather I do it in smaller chunks, like maybe 15-20 seconds worth, then to the next. OI find that if you can keep it shorter and switch between them faster...the differences are more obvious.

So I still think that B is the Neutron mix...and to me, it was stuck out between A and C.

miroslav I hope you are not offended about what I told you.

I don't get easily offended. This is all just entertainment, even when we are strongly disagreeing. :)

I am curious what your real goal is with this.
OK...so you want to see if people can guess the Neutron mix, and/or say which one they liked better...but is this something to do with specifically promoting Neutron as "good enough" to use instead of doing your own mixes?
Or are you testing your own skills against it...?

What are you looking to achieve with this survey?
 
Boy, if I had to mix this kind of music, I'd have Dr Kevorkian's office on speed dial. (Sorry for you lovers of this stuff.)

And, there's 1/2 hour of my life I won't have back. I truly don't know what Neutron would do with that, and never having used it, have no idea how you would "program" it to make mix decisions. For me, gritting my teeth through 2 listens, first the full pieces on one set of good cans, and then a second on my studio monitors and another set of cans (1/2-1/2), they really seemed to go from bad-better-best for me, but primarily because of the intelligibility of the vocals. I had a hard time believing a human could have mixed the first two, but, as I implied, this is not my thing and maybe they were ideal...

I think, as suggested, this is a little crazy kind of test, since even if the humans have a studio, it doesn't mean anything if they're not adept in this kind of material, or any other number of factors, including how much tweaking Neutron got - I mean, was it as simple as pick a genre and dump in the stems and take what comes out, or did someone get multiple passes over it, twisting knobs a little each time?

P.S. And again, the genre of "music" here is meaningless for me to make a decision about considering using Neutron, regardless of poll outcome. If anything, I'm disinclinded to consider it because I might think this is what it was designed for!
 
Boy, if I had to mix this kind of music, I'd have Dr Kevorkian's office on speed dial. (Sorry for you lovers of this stuff.)

And, there's 1/2 hour of my life I won't have back.

:laughings:


I truly don't know what Neutron would do with that, and never having used it, have no idea how you would "program" it to make mix decisions.

OK...since you asked...here's some general info on how it does things. ;)

First, I'm curious if the Neutron mixes for the clips were all from the app's algorithms, left at face value, and derived from the quick analysis of the audio that it does...or was there any amount of human tweaking of the application's chosen processes and settings...?
See...even if you let Neutron analyze and make the mixing decisions, there are initial "presets" that the user must choose based on the music type/style/etc you think the song falls into. So even if you then let Neutron do the rest, those initial preset choices can make a significant difference.
IOW...not everyone might chose the same initial presets for the same song.

I just want to make the clarification that it's not truly 100% all done by Neutron. The user provides some initial input before the engine analysis of the audio.You can also tweak as much as you want, like with any other plug.
That said..I still feel that regardless of user selection for the initial category/style that you want Neutron to use, I noticed that once Neutron does its thing...it tends to hype up the mids/highs always to some degree.

When it first came out a year ago or so, I checked out the demo, and I also heard example mixes that were provided by (? not sure who did them). I only messed with the demo for a couple of days and came to the conclusion that I never much liked what Neutron did regardless of those initial category/style selections made. I always wanted to change things.
After it does its analysis of the audio and spits out its mixing decision...there's just too much (IMO) shit added to the signal path. Like almost always there's lots of compression, and almost always there's a sonic/harmonic enhancer added (which is why I think there's that common hype in the mids/highs).

Of course...once you get the Neutron "decision" after analysis...you can start tweaking it and take stuff away or add stuff, so you can "override" what Neutron picked...but honestly, once you get to that point, it's not much of a different process than what you do using your regular assortment of plugs and tweaking them until you like the sound.
For the folks who simply don't want to tweak, or who want the application to make the decisions, because of the assumption that it's smarter than they are when it comes to audio decision making...I'm sure Neutron feels like a great thing to have. I just realized that I never really liked the Neutron decisions at face value...so then why bother with all that. I can get to the same place with my existing mixing tools and options.

That said...it does package a bunch of tools all into one app, which is probably another selling point for the guys who don't already have their tool set...but TBH, I think after you mix a lot of stuff with it and let it run at "face value" for the most part, you might notice that all your stuff has a "Neutron sound"...which may or many not appeal to you.
I think that is inevitable, considering that Neutron runs off of presets and coded algorithms.
While it can make initial adjustments based on the audio signal it's reading, it doesn't have the ability to actually "hear" the audio...and that, IMO is why you will end up with a "Neutron sound" on your mixes. There's only so many ways it can go...without you cutting in and making a lot of manual tweaks....kinda like what we do already. :)

So yeah...I didn't mention all this earlier, but that's why I already had/have an opinion about Neutron, and why at first I didn't bother to even listen to the three clips.

After demoing Neutron for a couple of days last year...I knew it wasn't for me, both for the overall sound of it at face value settings, and because I already have plenty of plugs to work with...but I can see why it would appeal to some people.
 
I've been using Neutron for the past year and I think it's a great tool. A couple things about Izotope Neutron...

First of all, it doesn't "mix" your tracks. Neutron is a channel strip that has compressors, a limiter, EQ, transient processing and an exciter in a single plug-in. You manually need to mix your tracks setting panning and levels.

The "track assistant" will sample a short clip of a given track after you have picked a general style for the track and then set up the compressors, EQ, etc. for the track. Izotope states that this is a STARTING POINT to work from. I find it useful for picking out frequencies I might want to EQ and just like a preset on a compressor, finding a starting point, but I usually make a lot of changes. It is still very helpful. It wasn't intended to totally automate your settings although sme people will be lazy and use it that way.

There are a lot of nice features like dynamic EQ, and the transient shaper saved my neck on a drum track with a bad sounding snare and a crash cymbal that originally sounded like a garbage can lid.

However, if you have a great set of plug-ins that you already know how to use, why would you want to change your entire workflow?

It will be interesting to see the results of the survey. In all honesty, I wasn't impressed with any of the three mixes.
 
First of all, it doesn't "mix" your tracks.

You're absolutely right. :thumbs up:
I think we all use the term "mixing" rather loosely, and it's as much a description of the processing done to individual tracks as it is about actual mixing/combining of the individual tracks.

Do you use it just on your individual tracks...or also on the stereo mix?
That too isn't clear about the survey mixes...where/how exactly was Neutron used.

When I demoed it, I thought it might have some application as "guide" for which way to go with mixes...but for me, it just didn't really provide anything I would want to use. I was hearing the mixes differently (for right or wrong :D) in my head than what Neutron was spitting out.
 
"How many guys does it take to mix a song?
1 to mix it, 999 to tell you how they would've done it, and 1 plugin to show how AI would do it" :facepalm:
 
I was under the impression the software 'did' the mix. Perhaps I misinterpreted the OP.

OP, it would be worth clarifying :

[COLOR=rgba(0, 0, 0, 0.870588)]"Two of the three examples were produced by two different mix engineers while the third one is the mix resulting from automated Neutron processing."

[/COLOR]Who created the third mix? An AI? A human, but based on tools chosen by an AI?
 
Back
Top