When will we get an A.I. DAW?

[TheLegit]

Say you've recorded a bass track, stick it on a bass track preset it will automatically apply all relevant compression, filtering EQ resonances etc, and then compare to say the kick and do any required ducking or EQ'ing to make the bass and kick go together.

This A.I. will do everything required to other tracks in the same way (all the programmed actions/ presets set up by the manufacturer to closely follow a template will deal with vocals, guitars etc, EQ to ensure clashing frequencies are dealt with, vocal widening, tuning, timing), leaving you with an A.I. mixed track. You can then go in and adjust to taste! You can have presets for Metal, Rock, Pop, EDM etc

I personally think it's a matter of time maybe five years?

[Guest]

I am not sure. I don't think the computer will ever mix a track a bad as me. why

[Drew Stephenson]

If you look at what Izotope are doing already, that's clearly on the cards.
Of course that will only give you a vanilla mix*, but a) that's already going to be better than a lot of bedroom producers (probably including me) can achieve, and b) it'll be a great time-saver and starting point to then go on and create something more interesting/individual.

* For a lot of purposes, that'll be all that's needed.

[Hugh Robjohns]

It's certainly on the cards. And, although not as involved, that kind of preset processing is already standard for photography in smart phones etc.

I'm not sure about your timescale. I think 5 years is a bit optimistic, but within a decade should be realistic.

The determining factor is going to be whether there would be a market for a 'self mixing' system. What proportion of recording musicians would want to invest (heavily -- that R&D wouldn't come cheap!) in something like that?

[Albatross]

I wouldn't want it, would take all the fun out of it. And all those happy accidents would be gone. Everything would just end up sounding the same. As Hugh said, its already there with photography and those facebook filters make everyone look like they fell out of that movie 'The Surrogates'.

Might be good though because as listeners are always looking for something different and the 'authentic' I'll be able to press my old 1960s Philips mono reel-to-reel into action and do some electric folk tunes while the lazy loopers are wallpapering the world with digicrap.

[RichardT]

I am sure it will happen bit by bit over the next 5-10 years. It will take some of the drudgery out of mixing, I hope.

[beej65]

For this relative newbie... would it be fair to say that an excellent rough mix (balance and panning only, no fx at all) would be that same starting point? Or is it better tothink of that stage as a pre-mix before, say, our current state of the art AI DAW's intervention (Izotope was alluded above, I assume referring to the various Assistants in Neutron/Nectar/Ozone etc.). And would any have any other examples beside Izotope of where similar AI might be found?

[Folderol]

I agree with Albatross. For me it would feel like trying to dance in a diving suit

[The Elf]

"This ain't no technological breakdown. Oh no, this is the road to hell..."

[BJG145]

Interesting idea, and I'm sure it's possible, but I'm not sure if it's mass-market enough to take off. As Hugh says, the automatic image processing built into smartphones nowadays is remarkable, but everyone wants to take photos; not everyone wants to mix tracks. And the people who do, don't want to do it like that, at least, not yet.

[Sam Spoons]

The only people who could materially benefit from such AI would be the pros who might save some time but they probably do a better job just as quickly as any computer what is the point. For most of us it would defeat the object of the exercise, the pleasure derived from doing it yourself (however imperfectly).

[RichardT]

The only people who could materially benefit from such AI would be the pros who might save some time but they probably do a better job just as quickly as any computer what is the point. For most of us it would defeat the object of the exercise, the pleasure derived from doing it yourself (however imperfectly).

For me, the final stretch of getting something ready for release is a pure slog - if AI can help with any of that, I’d welcome it!

[Drew Stephenson]

The only people who could materially benefit from such AI would be the pros who might save some time but they probably do a better job just as quickly as any computer what is the point. For most of us it would defeat the object of the exercise, the pleasure derived from doing it yourself (however imperfectly).

I'm not convinced there Sam, i think there are a lot of musicians who aren't interested in the engineering side at all and just want to create a product. I suspect that's a larger potential user base than people who want to understand the difference between a dynamic eq and a multiband compressor.

[Watchmaker]

When will AI be Picasso, or Bob Clearmountain? Will AI read Neruda poems on a stormy night and feel the wrench in its virtual gut?

I can't wait for AI to eat my breakfast! Is there nothing that computers cannot do that is not worth doing ? Why have art or interest in anything at all?

That is the land of death and I hope the acolytes find it enjoyable. One thing I can promise you, exactly no one will be inspired by an AI mix.

[Drew Stephenson]

Hmmm. I suspect most people are more interested in the song than the mix.

[ManFromGlass]

from the economic point of view -
Your personal Song AI creates 1000 songs today before lunch and gets them on the net to the appropriate genre sites.
Each play or partial play nets you .oo341 of a cent.
You’ve been creating 1000 songs a day for a few years.
You also own the publishing to all (have to switch to calculator for a sec) 1,825,000 plus tunes. Times this by the number of plays and the .00341
You might be able to afford food for a week by the time this all happens but mostly because of the cost of food.

[Rob Hansen]

Lol! This question could logically be phrased in a way that is certain to upset some people: "When do I get an A.I. engineer?" I think the results would be interesting, like cgi characters in movie. That idea seemed to threaten actors for a time but ultimately turned out to be a moot point. Similarly, good human engineers need not fret. A recording A.I. would be okay for a basic assist, but I'd really just rather have something to talk to.

[The Elf]

Hmmm. I suspect most people are more interested in the song than the mix.

I'd agree up to a point, but let's take an example... Buggles' 'Video Killed the Radio Star'. How would a 'normal' mix of that sound?

I don't want to live in a world where that kind of creativity is deferred to the 'beige everywhere' world of AI.

[Wonks]

Technically, I can see the AI would only work with a limited number of plug-ins otherwise it won’t know how to control them. Yes, you could have a learn mode so you can map say five of the most common features of that type of plug-in, but if the plug-ins got twenty parameters to adjust, then there’s a whole lot of stuff that makes that plug-in special going to waste.

So you end up with every track using the same limited range of available effects. OK, that didn’t stop great music being produced in studios in the pre-DAW age where you just had the studio’s rack effects (plus maybe a plate reverb or echo chamber), but it is a big step backwards.

On the other hand, if it allows people more time to concentrate on the performance and getting the sound right at source etc. then that could be a good thing.

[ore_terra]

I can think in a few commercial genres that wouldnt suffer too much with an automated mixing process... in the end most tracks sound exactly the same, with same effects, same balancing, same... all!

I know a producer/mixer here with quite a few reggeton hits who always use the same template, the same libraries and the same 2 or 3 singers. once the song is programmed and recorded he says it takes no more than 15 mins to get it mixed and mastered, and that includes a bit of editing time too

[Drew Stephenson]

Hmmm. I suspect most people are more interested in the song than the mix.

I'd agree up to a point, but let's take an example... Buggles' 'Video Killed the Radio Star'. How would a 'normal' mix of that sound?

I don't want to live in a world where that kind of creativity is deferred to the 'beige everywhere' world of AI.

This is why I'm not concerned about it. But it is coming and will be enough for a lot of people.
Look at it this way perhaps:
Take a random scroll through SoundCloud, consider how basic / bad some of the mixing is there. Stuff that's obviously been recorded in an untreated room, with no understanding of mic positioning, loads of low frequency rumble because the user doesn't know what a HPF is, lots of lisping because they've discovered the de-esser but have completely over-cranked it, the whole thing is flat as a pancake because they stuck a limiter on the master buss and cranked it to make it loud... etc etc etc.
These are the songs and the people for which it'll be a godsend.
And actually it'll be a benefit for all of us because all of these tracks will end up better and we'll be able to appreciate the songs more.

What it will do though is bring the kind of conversations that mastering engineers put up with down to the mix engineer as well.
You know the ones I'm talking about...

[worshiptuned]

Or is it better tothink of that stage as a pre-mix before, say, our current state of the art AI DAW's intervention (Izotope was alluded above, I assume referring to the various Assistants in Neutron/Nectar/Ozone etc.). And would any have any other examples beside Izotope of where similar AI might be found?

Yes, EZ mix 2 of Toontrack. I didn't test it. But I have Izotope and I was quite excited to try the Izotope Assistants, the mix assistant, the vocal assistant... The plugins sound good, but the assistants...it was so disappointing. I thought they were a good starting point and would lift my productions. After a few trials I ended up thinking that my very modest mixing skills provide a much better starting point! I would suggest the assistants for people that don't know absolutely anything about mixing.
I found useful instead the tonal balance control that helps to compare your mix with the average tonal balance of other mixes of the same genre.

[ManFromGlass]

I was disappointed too. A digital assistant sounded cool.
I think with the AI there would be choices of “In the style of” or even better “In The Virtual Studio of”
And poof, there would be a CGI Bob Clearmountain or Sir George Martin greeting you and doing your mix.

[Hugh Robjohns]

I think Drew is right; AI mixing wouldn't be a replacement for real engineers, and wouldn't challenge the creativity and talents of a good engineer. And it probably wouldn't appeal to anyone with a genuine interest in the skills and art of mixing...

But it would be useful and I would appeal to musicians who just want something to deal with building a basic balance and dealing with 'all that technical stuff' that holds n interest for them. And even if it did no better than create acceptable demo mixes, that would be seen as a great thing to that market.

The only question is whether that market is sufficiently large to warrant the R&D and whether it would be willin to pay enough for an AI mixing app to make it commerciall viable.

I have my doubts....

[RichardT]

Sampled instruments haven’t completely replaced real instruments and musicians.

Automated manufacturing hasn’t completely replaced craftspeople.

Camera phones haven’t completely replaced professional photographers.

It will be the same here. It will enable more people to produce reasonable tracks, but it won’t replace the skills required to produce the best stuff.