How are Dolby Atmos mixes created?

[DWG]

I'm curious if there is somebody here that can answer my question. This pertains to music only (not movies). I understand how 5.1 mixes are created. A guy like Steven Wilson obtains the original multitrack tapes (or files) loads them into the mixing console and meticulously determines what sounds go into what channels. Sound a bit time consuming. I see he is now mixing his multichannel in Atmos. My question doesnt really pertain to his work, rather it pertains to what I am hearing on Apple Music. I cant tell you how many titles are available in Atmos - it seems like thousands. If the process of making an Atmos is manual, how are so many titles getting created so quickly and who is creating them? The alternative it seems is the possibility that some piece of software is taking the stereo master and upmixing to Atmos? I hope thats not the case.

 

[krabapple]

Some are upmixed from stereo, for sure..

But I too am curious to know what factory-like setup is cranking out all these streaming Atmos remixes.

I've asked over on Quadraphonicquad, might be worthwhile to keep an eye on responses there

 

[Oddball]

That is a very good question. And I can't answer it but hope someone else with more insight could.

For movie soundtracks there is usually a mixing engineer involved and they do a better or worse job. Some remixes, like Patriot, are flat out fantastic, some less so. But movie titles are bigger ticket/value items so that makes sense. Also not sure to what extent they do or not use some algorithm to create a base version for the mixing engineer.

For audio tracks, there are literally millions of titles. There are up mixers like Dolby, DTS, Auro that will do it for you out of 2.0 track. To what extent they use it in studios (or some different algorithm) is the key question. I would imagine there is generally less of human intervention in these "bulk" titles, but some might receive full human atention.

BTW, a good Atmos mix will not be channel based, but object based. So that it will play well on 5.1.2 system as well as on 11.8.8 system. Sound is not mapped to a particular channel, but to a particular place in the Atmos bubble. Atmos decoder will then place it to the right channel/s based on the capabilities of the reproduction system.

 

[krabapple]

Yes, but most of that kind of misses the point.

How, logistically, is this cornucopia of new Atmos music remixes being generated every week for streaming?

Auto-upmixing from stereo sources would be one way, but that does not obviously seem to be the case. (And would be rather scandalous if it proved to be generally true.)

The alternative is a flood of multichannel files (whose own creation is a black box) being provided to an army (of unknown size) of human mixers, who are then creating Atmos mixes from them.

 

[EERecordist]

There is software. I would hope there is a human in the loop. Some works with a budget are manually remastered. You can find interviews by the engineers manually remastering The Rise and Fall of Ziggy Stardust by David Bowie. They had the stems/tracks and they put each separately through processing including room echo on the drums.

This came across my feeds a few days ago.

 

[holdingpants01]

It's very quick if stereo version is already finished. You get the final mixes split into the stems or tracks and just put them around, if you have time you can move them using automation or use real multichannel reverb instead of upmixed stereo stem and that's it. You don't need to touch EQ, you usually don't even use compression as those are baked in already. Doing it from the scratch is a whole other thing, but it rarely happens. One could do the whole simple album in one day, but probably won't admit it as it's bad for business. Bigger name producers and mixers have assistants that do the "transfers" for them, so the client gets both versions basically at the same time. It's more complicated when dealing with old material, for example Thriller was mixed from the scratch in stereo to be similar to original, then the stems were created and transferred to atmos, but you can't listen to the new stereo version, you get original mixes in stereo and new ones in atmos

 

[krabapple]

For legacy tracks the stereo versions may have been finished decades ago.
So there has to be recourse back to their multitracks, unless it's simply software/AI splitting out parts from the legacy stereo mix.

The question is about volume. How so many are generated so fast.

'Just put them (stems/tracks) around' in the sound field seems to glide past mix decisions that actually involve humans, which takes time. Unless you're saying they usually don't?

 

[holdingpants01]

'Just put them around' seems to glide past mix decisions that actually involve humans. Unless you're saying they usually don't?

like I said if someone is working on a current releases, all the mix decisions were already made and are printed into the stems, so there's really mostly placing thing around involved. For old multitracks that's more complicated, and for old stereo mixes if there's no multitrack it's quite easy, you just slap the reverb or whatever the label accepts, it will sound bad anyway and they usually can't even hear it properly without dedicated atmos room

 

[holdingpants01]

The question is about volume. How so many are generated so fast.

there's are quite a lot of people that can do it (you can basically do that using only Logic with headphones) and new pop music is delivered in atmos almost as a default, at least in the US. I dismantled my atmos setup in my studio as no one wanted to pay more for it in the niche I'm in. I'm doing infinitely more vinyl and magnetic cassette deliveries than I did atmos for the last 1,5 years. At least it's in stereo most of the time!

 

[krabapple]

like I said if someone is working on a current releases,

Except you did not say that.

all the mix decisions were already made and are printed into the stems, so there's really mostly placing thing around involved.

Ok, like I'm 5 years old: by 'already printed into the stems' do you mean that their *placing around* the Atmos sound field are already specified? There's no decisions involved when rendered into Atmos?

For old multitracks that's more complicated, and for old stereo mixes if there's no multitrack it's quite easy,you just slap the reverb or whatever the label accepts, it will sound bad anyway and they usually can't even hear it properly without dedicated atmos room

Well again, you're moving sideways from what I'm asking. Yes, if the only thing available is the stereo mix, some sort of upmix/fakery is necessarily the only way to go.

But I am hearing lots of quite old content on Apple Music where it sounds like rather more than that , e.g., "I Feel Love" by Donna Summer/Moroder. Granting that upmixing, by itself, when done well, can be impressive (as I know from simply doing it at home)

 

[holdingpants01]

Ok, like I'm 5 years old: by 'already printed into the stems' do you mean that their *placing around* the Atmos sound field are already specified? There's no decisions involved when rendered into Atmos?

You print mono and stereo stems and place them in the atmos bed or as an object, preferably there's no change of volume, EQ or compression, it's all printed in but the placement. Some adjustments are made, but with stereo reference file in the session it's pretty straightforward. The most important thing is to meet the volume and length guidelines, but they're purely technical

Well again, you're moving sideways from what I'm asking. Yes, if the only thing available is the stereo mix, some sort of upmix/fakery is necessarily the only way to go.

But I am hearing lots of quite old content on Apple Music where it sounds like rather more than that , e.g., "I Feel Love" by Donna Summer/Moroder. Granting that upmixing, by itself, when done well, can be impressive (as I know from simply doing it at home)

Yes, it's more complicated on old material and there's no one standard way to do it, as all material comes in different ways. You can have only the master tape in stereo and with that you can try to use AI stem separation to have some room for creativity, or just slap the reverb. You can have tape multitracks before or after the mixing. You can have stems that were created for 5.1 releases, or for old games like Guitar Hero. They all can come in digital or analogue formats, there could be some work involved in transferring them, or you can have a pro tools session with everything working from the start. Then there's a budget or there isn't one, if a label wants to invest in an atmos release they will hire some big shot to do it, if there's none and they just need to meet the atmos quota they can give it to a random dude with a laptop. Hard to say what was done in Donna Summer song, but for such big acts I'm pretty sure they didn't skimp on doing it properly

 

[goat76]

Ok, like I'm 5 years old: by 'already printed into the stems' do you mean that their *placing around* the Atmos sound field are already specified? There's no decisions involved when rendered into Atmos?

He means that a "stem" is usually a track containing one of the instruments in a song that is already processed with EQ, compression, and everything else needed to fit into the mix. The "stem" can be a mono file or it can be a stereo file of one of the instruments in the stereo mix, but as this "stem" is already processed it can easily be used to create a separate Atmos mix fairly quickly as the more time-consuming work of how those individual instruments sound is already done for the stereo version of the mix.

Well again, you're moving sideways from what I'm asking. Yes, if the only thing available is the stereo mix, some sort of upmix/fakery is necessarily the only way to go.

But I am hearing lots of quite old content on Apple Music where it sounds like rather more than that , e.g., "I Feel Love" by Donna Summer/Moroder. Granting that upmixing, by itself, when done well, can be impressive (as I know from simply doing it at home)

You have probably noticed that most old records that are made for Atmos are mostly the work of big and famous artists, and most of them were big and famous already when those records were made, so the multi-tracks of these recordings were treasured as "gold" and likely digitized at a later stage to make sure the quality didn't degrade over time. In the analog time of age, the processed (as far as that was possible) "sound" was usually printed to tape so these old recordings didn't have to be too complicated to be mixed to Atmos. But sometimes when there isn't a separately recorded room sound to be found on those old tapes (or on the digitized dito), it's fairly common that the Atmos mixing engineers will create a new room sound by playing the recording in a live room in the studio or sometimes adding an artificial reverb to the Atmos mix.

EDIT: I see that @holdingpants01 gave a similar answer while I was slowly typing my response.

 

[kemmler3D]

The question is about volume. How so many are generated so fast.

As @holdingpants01 says I think if you already have the stereo mix and the stems (i.e. separate drum, vocal, guitar, etc. tracks) you could crank out one album per day, manually. The hard work of balancing tonality and dynamics is done already, so you're just moving things around in space.

In fact if you were being lazy, you might just swap out stereo reverb for 3D reverb and call it a day. In which case an "atmos" mix could be done in maybe an hour per album.

So, take 1x per day or maybe 5x per day and multiply that by the number of people who know how to do this, which is probably in the thousands, and you have a plausible explanation for how this many are getting done.

 

[goat76]

I find it highly unlikely that there are many Atmos mixes not made out of the original multi-tracks or multi-stems, well, unless there is obviously just the same mix as the stereo mix with just some reverb added to the surround channels. If that is done it would be pretty easy to hear.

 

[kemmler3D]

I find it highly unlikely that there are many Atmos mixes not made out of the original multi-tracks or multi-stems, well, unless there is obviously just the same mix as the stereo mix with just some reverb added to the surround channels. If that is done it would be pretty easy to hear.

Yes, I think you would only do that if you only had access to the stereo mix and it wasn't a very good album.

 

[krabapple]

He means that a "stem" is usually a track containing one of the instruments in a song that is already processed with EQ, compression, and everything else needed to fit into the mix. The "stem" can be a mono file or it can be a stereo file of one of the instruments in the stereo mix, but as this "stem" is already processed it can easily be used to create a separate Atmos mix fairly quickly as the more time-consuming work of how those individual instruments sound is already done for the stereo version of the mix.

OK, let's confine ourselves here to new releases where stereo and Atmos are done in tandem, not legacy that was originally analog:

I do understand what a stem is, thanks. Yes, an element's EQ, compression can be set by what it is for the stereo mix (but I don't know if he/you are also claiming that the level of the object in space relative to all the others is also already set:. Thus no adjustment necessary when remixing to Atmos. Does that really work when, eg. the object is put in the height speakers?)

The primary way a surround mix differs from a stereo mix is the placement of sounds in space. For stereo, you pan elements left/right/center (unless its 'live' and purist). It's rather more complex for Atmos, no? What really smacks the listener in the head is the surround sound part. That decision -- where each object goes -- is crucial, and that's what I'm talking about.

You have probably noticed that most old records that are made for Atmos are mostly the work of big and famous artists, and most of them were big and famous already when those records were made, so the multi-tracks of these recordings were treasured as "gold" and likely digitized at a later stage to make sure the quality didn't degrade over time. In the analog time of age, the processed (as far as that was possible) "sound" was usually printed to tape so these old recordings didn't have to be too complicated to be mixed to Atmos.

And now we're talking legacy audio.

Were/are Digible Planets 'big and famous' artists? They had a brief moment back in 1993 or so.
Right now there's a stunning Atmos mix of their debut album (and I'm told, their far worse-selling 2nd one), streaming on Apple Music. I couldn't believe it when I stumbled across it in a click-through.

Rinse and repeat for lots of artists/albums/tracks that weren't really 'big and famous', in addition to lots who were.

But sometimes when there isn't a separately recorded room sound to be found on those old tapes (or on the digitized dito), it's fairly common that the Atmos mixing engineers will create a new room sound by playing the recording in a live room in the studio or sometimes adding an artificial reverb to the Atmos mix.

Have you actually surveyed/listened to the Atmos 'Spatial Audio' offerings of legacy music on, say, Apple Music?

I'm really, really, not talking about 'room sound' (actual or simply:synthetic reverb) added to mixes, though those sorts of releases exist, just as they did for physical DVD-A/SACD releases (e.g., Miles Davis's 3-track recordings repurposed as '5.1' on SACD ages ago)

I'm talking about the many, many, many Atmos remixes of legacy audio that seem to have required someone to, at minimum, decide where to 'pan' things in 3D space.

How 'fairly quickly' can this be done? How many people are involved in this effort? It seems massive, to me.

 

[krabapple]

As @holdingpants01 says I think if you already have the stereo mix and the stems (i.e. separate drum, vocal, guitar, etc. tracks) you could crank out one album per day, manually. The hard work of balancing tonality and dynamics is done already, so you're just moving things around in space.

In fact if you were being lazy, you might just swap out stereo reverb for 3D reverb and call it a day. In which case an "atmos" mix could be done in maybe an hour per album.

Have you actually sampled the offerings out there? Do you think this is what is actually, mainly, being done?

 

[kemmler3D]

Have you actually sampled the offerings out there? Do you think this is what is actually, mainly, being done?

I've actually never heard an Atmos mix but I know my way around a DAW... if the stereo mix already exists, translating it to Atmos should be fairly straightforward depending on your goals for the mix.

You could actually add more dynamic range and whatnot back into the mix in the process, but I would imagine a lot of times they just keep the compression more or less as it is and spread things out across more channels.

 

[goat76]

OK, let's confine ourselves here to new releases where stereo and Atmos are done in tandem, not legacy that was originally analog:

I do understand what a stem is, thanks. Yes, an element's EQ, compression can be set by what it is for the stereo mix (but I don't know if he/you are also claiming that the level of the object in space relative to all the others is also already set:. Thus no adjustment necessary when remixing to Atmos. Does that really work when, eg. the object is put in the height speakers?)

No, the level of the element/instrument will most likely be set differently in the Atmos mix as it will likely be perceived differently loudness-wise depending on where it is placed in the mix, so the level set in the stereo mix will most likely not be the same when a particular element/instrument is panned differently in the Atmos mix. But with that said, setting the level between different elements in a mix is usually one of the easier tasks in mixing, most of the work usually goes into carving out space for all the individual elements so as not to overlap each other too much frequency-wise, especially for a stereo mix which fastly gets "crowded" with a lot of frequency masking going on, which I have heard is less of a problem with an Atmos mix as the "canvas" is much larger as the elements can be spread out in a larger three-dimensional space.

The primary way a surround mix differs from a stereo mix is the placement of sounds in space. For stereo, you pan elements left/right/center (unless its 'live' and purist). It's rather more complex for Atmos, no? What really smacks the listener in the head is the surround sound part. That decision -- where each object goes -- is crucial, and that's what I'm talking about.

I'm not a professional mixer as I only mix the music I create on my own, but I have played a bit with mixing for 5.1 surround and it's not that much more complicated than making a 2-channel mix. It's the same play with panning, but instead of just having the space between the left and the right channels I can as easily place an element as a phantom sound wherever I want between the left front channel and the left surround channel so that the element appear to be coming from the side of the listener.

It's not much more complicated for Atmos, it's just that you now have more of a three-dimensional space in the mixing suite where you thanks to the ceiling speakers have the opportunity to "phantom" the position more freely in more directions. There are so-called "Beds" that are the core of an Atmos mix, but what sets Atmos apart from traditional surround mixes is the "object-based" nature where nothing is hard-set to any specific channels, and instead using algorithms to decide which speakers in a system need to "take the load" to place the element where it is supposed to be positioned in the three-dimensional space.

And now we're talking legacy audio.

Were/are Digible Planets 'big and famous' artists? They had a brief moment back in 1993 or so.
Right now there's a stunning Atmos mix of their debut album (and I'm told, their far worse-selling 2nd one), streaming on Apple Music. I couldn't believe it when I stumbled across it in a click-through.

Rinse and repeat for lots of artists/albums/tracks that weren't really 'big and famous'.

I only speak in general when I say "big and famous". There are of course many original analog multitrack tapes that are kept from recordings of both famous and smaller bands, and they can all be used to create Atmos mixes.

Have you actually surveyed/listened to the Atmos 'Spatial Audio' offerings of legacy music on, say, Apple Music?

I'm really, really, not talking about 'room sound' (actual or simply:synthetic reverb) added to mixes, though those sorts of releases exist, just as they did for physical DVD-A/SACD releases (e.g., Miles Davis's 3-track recordings repurposed as '5.1' on SACD ages ago)

I'm talking about the many, many, many Atmos remixes of legacy audio that seem to have required someone to, at minimum, decide where to 'pan' things in 3D space.

How 'fairly quickly' can this be done? How many people are involved in this effort?

I have Tidal which provides Atmos music, so I have listened to a fair deal they have in their catalog, and everything of what I've heard seems to be Atmos mixes made from the original multi-track recordings.

How quickly it can be done is dependent on every single recording, what type of music it is, and how the recordings were done. Some music like rock, jazz, and classical will probably not go that far from how the stereo mix is done, but it can still be highly complex even if it still sounds like all of the instruments are located in front of the listener. Professional mixing engineers are usually very fast knowing what type of mix will suit a particular type of music, they know when to keep it "traditional" and when they have free hands to make a more effect-like mix (that may take more time to do).

 

[Dogen]

Apple Music apparently does not accept Atmos tracks upmixed from stereo; they must be from the multitracks, or generated by playing the stereo track in a physical space and recording room ambience. So, I’m very interested too in how older recordings are rendered in Atmos…following this thread and the one on QuadraphonicQuad.