How much more overhead will running MIR in 5.1 take up?

[VirtualVirgin]

I run a large orchestral template with MIR pro in Nuendo.

I may have 70+ instances going at once simultaneously in stereo.

If I switch over to 5.1, how much more CPU and RAM can I expect MIR to eat up?

[Dietz]

If you're comparing a 5.1 MIR setup with its stereo version you will roughly see three times the CPU load, because you will need six convolutions for each source on the Venue's stage instead of two.  

RAM requirements will depend much more on the chosen settings, though, like the actual use of IR positions, Secondary Microphone and so on. It's safe to say that you will need about twice the RAM you would use for stereo.

A hint: It might be advisable to choose a 5.0 format instead of 5.1. You'll need 20% less CPU that way. There's not much MIR itself could contribute to the LFE (which is a dedicated Low Frequency Effects-channel, not a sub-woofer for bass management), due to the simple fact that the test signals used for capturing MIR's impulse responses start at 50 Hz anyway. :-) The 5.1 presets I provided are there for the sake of completeness, mostly.

For what it's worth: I'm mixing two hours of a movie score ATM, using about 60 5.0-instances of MIR Pro (plus hundreds of other audio tracks on a dedicated SSD). No active virtual instruments. The machine is old and quite modest by today's standards, sporting an i7-3970 processor from 2013 and 32 GB of RAM. I have chosen high latencies for both the audio system and MIR Pro's convolution engine and have yet to run into performance issues. :-) ... but keep in mind that MIR's option for Dynamic Processing allows for switching off any unused instance, so the arrangement of the piece at hand will define the limits of processing power, too.

EDIT: I'm working in 48 kHz. 

HTH,

[VirtualVirgin]

Hmmmm...

In that case I may be better off with keeping MIR in stereo and adding MIRacle for surround ambience.

That makes me wonder, does MIRacle have settings specifically for upmixing?

I am trying to figure out ways to mix into surround and TOA (Third-Order Ambisonics), yet keep my Virtual Instruments active from beginning to end in the process (I prefer to edit the source performance before I look to compressors and EQs to solve anything).

I really only need surround ambience for anything that is supposed to be on stage. I did try your trick with LCR in the conductor's seat and the 5.0 rears in the back row and it is nice, but that was only for a test with a few sounds.

I'm not sure my current CPU will handle a conversion for the entire orchestra as most of my instruments at this point are physical modeling (CPU intensive).

Also, you mention mixing a project now with MIR. Do you have any examples of releases done with MIR that we could hear? It would be interesting to hear what the developer does with his own tools!

[Dietz]

What about arranging in stereo and just adding the C and Ls/Rs channels for the actual mixdown? Just start in a 5.0 environment, stick to a L/R Main Mic for the time being, and instantiate a 5.0 Main Mix + Secondary Mic for mixdown only. You can do this as an offline bounce, so the CPU of your machine shouldn't be an issue. No need for artificial up-mixing. 😊

*****

Regarding some examples of my own MIR-based mixes: It always seems to be a bit strange for me to toot my own horn on these pages. 8-) I use MIR in pretty much all of my productions, either for the actual mix, or just for spatialisation of a few elements here and there. Some large-scale mix projects worth mentioning might be :

-> Borat Subsequent Moviefilm: Delivery of Prodigious Bribe to American Regime for Make Benefit Once Glorious Nation of Kazakhstan (2020) - original score, MIR used for all orchestral parts

-> The World of Hans Zimmer - A Symphonic Celebration - Live (2019) - album, MIR used in surround for the complete mix

-> Manaslu (2018) - original score, MIR used in surround for all orchestral parts

-> Conchita Wurst & Wiener Symphoniker: From Vienna with Love (2018) - album, MIR used for all orchestral parts

-> Life Guidance (2018) - original score, MIR used for all orchestral parts (BTW: The original mix was a first proof-of-concept of using MIR Pro in a true 3D-audio setup 😊 ...)

... and a bit older, but one of my favourite productions of the last decade:

-> JBBG: True Stories (2016) - album, MIR used for all brass and woodwinds, as well as many other elements like percussions and so on.

It might be worth noting that all of these examples are full-blown commercial projects, so there are of course many other tools used alongside MIR Pro (most notably additional algorithmic reverbs and delays), but I can assure you that it has been the foundation of all mixes mentioned above. :-)

[VirtualVirgin]

Interesting selections ;)  I like that MIR sounds good without having to be identifiable, in that it does its job well in sounding natural. It doesn't come off as an "effect" like a Lexicon or plate reverb. 

Question about the track Olfeder on the Manaslu soundtrack: are you using something like dearVR pro @ 1:08 when the synthesizers come in? The spatialization sounds quite different from a MIR soundstage.

"What about arranging in stereo and just adding the C and Ls/Rs channels for the actual mixdown? Just start in a 5.0 environment, stick to a L/R Main Mic for the time being, and instantiate a 5.0 Main Mix + Secondary Mic for mixdown only. "

I could, but as you walked me through in another post, it takes a bit of workaround for MIR to switch over to surround in Nuendo.  I considered switching my instruments over to VEP so I could also access the wet/dry mix functions, but I rely on Nuendo for channel disabling to run my template (so I can have hundreds of instruments ready to go and placed on the MIR stage). All I have to do is enable any instrument I need for my arrangement. 

"You can do this as an offline bounce, so the CPU of your machine shouldn't be an issue. No need for artificial up-mixing. :-)"

Like I said, I prefer to have access to MIDI data all through the mix process so I can tailor dynamics and articulation. Occasionally I do render things, but the switch-over of 70+ tracks of MIR over to 5.0 is real speed-bump in the whole practice. I think I will wait until the reviews are in on the next rollout of AMD Threadripper processors (possibly Q1 2021) to see if they have solved the latency issues for audio work, then setup a template with all MIR outputs going to Ambisonics channels in anticipation of the next MIR release ;)