Mixing orchestral music in Atmos

muk said:

I am not clear how the routing in an Atmos template is supposed to work to facilitate the same stems. As everything is routed through the Dolby Renderer, is there a way to set stems in the renderer itself? Or do I have to create group tracks as stems before the renderer?

When searching for answers all I came up is information about how to use stereo stems to create an Atmos mix. Is there no such thing as Atmos stems?

Click to expand...

There are ways of doing some of this, but they're painful. Atmos is more like a Master format, rather than a tool that makes the constituent parts easily accessible and exportable.

If you only need to export a few combined stems for a dub, then there is a Grouping functionality available in the renderer - this is designed for post houses to make their DX, MX, FX stem deliverables from an atmos file, but you can re-label and add your own groups. You would create a separate bed per stem, and group it with any individual objects that also comprise part of that stem. Once you've printed your track into the renderer, you can go in and export Re-renders for each group in whatever channel based format you need - 5.1 or stereo most likely.

The other alternative is to route to stems in your DAW, prior to routing out to the Renderer.
This is more or less the route I take when mixing films in Atmos. Each stem comprises a 7.1 width track that then routes to the Atmos bed, 2x stereo tracks which route to objects as overhead channels and a further stereo track for anything that needs to fly around (v rare) or just positioned as a static width object. These all get printed as audio tracks before routing out to the renderer. I use sends off these audio tracks to create a quick stereo downmix as a composer listening reference. The printed audio tracks are the deliverables that get sent to the dub where they can easily deal with them without having to worry about combining complex object allocations or metadata going awry.

Working like this creates a lot of system overhead. It may not even be doable depending on which DAW you're using (PT handles it pretty well though).

TLDR: It's not easy to create simultaneous Atmos, stereo and stem mixes in a one-stage process. The format can be really cool, but play with it extensively before deciding whether it's actually practical and beneficial to you in your professional reality.

Adam, I really appreciate your input. Thank you!

If I understand you correctly, delivering stems in Atmos for music alone is uncommon. The stems are Dialogue, Music, Effects etc., rather than sub-stems within the music. That's good to know.

From the two possibilities you mentioned, the first one sound simpler in terms of routing. As long as I create a bed (and objects as needed) for each section of the orchestra, I can then branch out and render the stems as needed. That sounds reasonable and manageable! I am definitely going to try this! It sounds like a logical way to structure the mix template in any case. Great tip.

Routing the stems in the DAW prior to the Renderer is basically what I have for my stereo mix. Why do you branch out the height channels individually? It's true that Atmos stupidly allows only two height channels for beds. But you could create 'Object Beds' with a channel width of 7.1.4, or whatever you need. Why didn't you choose that way? Wouldn't it be simpler in terms of routing?

For the stereo mix, I also print the tracks to audio before mixing. I will try to implement something similar for the Atmos mix. I hope that's possible when grouping the stems only in the Renderer.

I will make some test to see whether my pc can handle the system load with Atmos. For large orchestral projects it might be difficult. I will see. I hope there are workarounds like freezing tracks etc.

muk said:

Adam, I really appreciate your input. Thank you!

If I understand you correctly, delivering stems in Atmos for music alone is uncommon. The stems are Dialogue, Music, Effects etc., rather than sub-stems within the music. That's good to know.

Click to expand...

Kind of - these are deliverables that the dub will ultimately create at the end of the project. In the interim, someone like me will still be delivering music sub-stems to the dub.

The idea of delivering music as 7.1 beds and stereo overhead objects is that it gives the dub relevant spatial content to work with, but can also easily be re-panned/re-mapped if needs be. I'll also not deliver overheads/extra stereo objects for every stem, only where the content makes sense such as an orchestra that's recorded with the real height information.

muk said:

Routing the stems in the DAW prior to the Renderer is basically what I have for my stereo mix. Why do you branch out the height channels individually? It's true that Atmos stupidly allows only two height channels for beds. But you could create 'Object Beds' with a channel width of 7.1.4, or whatever you need. Why didn't you choose that way? Wouldn't it be simpler in terms of routing?

Click to expand...

It's only quite recently that Protools allowed the creation of internal 7.1.4 and above busses so haven't integrated that into my workflow yet, but yes there's no reason not to do as you've suggested with Object Beds. You just need a daw that will allow you to pan freely around a 7.1.4 bus. The use of height information in Atmos for film music is generally quite conservative though, typically the orchestra recordings will have discrete height mics recorded that will be panned to the overhead objects, but I don't spend too long finessing height information on synths, programming, perc etc. Leave that to the dub.

The information is gold. Thank you Adam. I feel better prepared to tackle the template update. I will dive in head first and I am sure I will have more questions soon.

I have read up on Atmos, and watched some videos on the workflow. But I have yet to try it, just haven’t had any requests for it so far. So, going slow here.

One thing that can't be pointed out often enough is that ATMOS as a format was _not_ invented with music in mind. As great as the whole concept is for scalable "audio for media", with all-important intelligibility of dialogue and such, the idea of having completely separated tracks a.k.a. "Objects" is quite alien to music production with delicate balances and timbres. It also stands in stark contrast to all our habits and workflows to achieve "glue" and coherence.

My personal approach is to mix in discrete 7.1.4 and to only encode the final product to ATMOS (if needed at all) with a 7.1 bed for the main layer and four fixed Objects as virtual speakers for the top layer. All modern DAWs are able to pan in 7.1.4 natively. Down-mixing to 5.1 or stereo is a straight forward task. And we can implement "broadened" versions of our stereo bus processing, too.

May I recommend checking out the Mixing With Mike (mixingwithmike.com) Immersive Mixing course when parts II and III are out?

The part I that's currently available covers creating immersive-type effects on headphones/2 channel speakers, but the parts II and III (still to come out) will cover Atmos and Binaural mixing.

I've watched a lot of his material and it's always thorough and useful.