Loopback / cable-free audio routing for Mac

[Piet De Ridder]

Rogue Amoeba Loopback allows you to create virtual audio devices in a Mac, enabling you to take the sound from any application and/or audio input device, and send it to audio processing applications. Loopback, they say, gives you the power of "a high-end studio mixing board" right inside your Mac.

Some uses for Loopback:

Combine Hardware Devices - Apps like GarageBand, Logic, and Ableton Live only record from a single audio device at once. Thankfully, they offer recording from many channels. With Loopback, you can combine multiple input devices into one virtual device for easy recording.
Screencasts - Screen recorders, including QuickTime Player, allow you to include either microphone audio or all system audio at once. Neither option is ideal. Create a virtual device that grabs just the mic and the app’s audio to get exactly the audio you want.
Play Audio to Podcast Guests - Combine your mic with audio sources like iTunes or QuickTime Player, then select your Loopback device as your source in Skype. Presto! Your guests all hear both your voice and your audio add-ons.

Here's the link: https://rogueamoeba.com/loopback/
And here's the Loopback blog: http://weblog.rogueamoeba.com/2016/01/1 ... -loopback/
There's a free download to test it, buying it means $75.

[Marius]

A fantastic app, glad it's finally out. Rogue Amoeba are a wonderful company and if anyone was to tackle this particular issue, I'm glad it was them. For anyone doing podcasting, screencasting, etc. It's a wonderful improvement over the existing solutions.

[Aesthete]

Would this let me route itunes into Logic for processing ?

[tack]

I heavily rely on the loopback feature of my RME interface (via TotalMix). Very cool providing this in an interface agnostic way, and in a way that can span multiple interfaces. Too bad it's Mac only.

[Piet De Ridder]

Would this let me route itunes into Logic for processing ?

I think so.

What I'm most curious about, and which I'm going to try as soon as I can, is if it'll allow me to route audio out of Logic and into SPAT (standalone) and then back into Logic. Multi-channel out, stereo in, of course. That would change the (SPAT-)game enormously.

_

[Lawrence]

Allowing you to use SPAT resources more frugally?

[Piet De Ridder]

Not just that. But bringing up to eight sources into one single SPAT-space and having these sources, all of them together, influence the room response for the whole group. That's fundamentally different from giving each source its own SPAT treatment.
A room responds differently to a bass than it does to a flute. It'll also respond differently to the combination bass+flute. And SPAT can do that with up to 8 sources.

_

[Piet De Ridder]

Isn't going to work. SPAT, being an AU-plugin, needs to run in some host or other, so what I need now is a (preferably simple) AU-host, to run alongside Logic, and that can receive 8 stereo-channels (and send out a single stereo channel). I thought I had the very app in DSP-Quattro, but I can't get it configured the way it needs to, and for some reason it also refuses to see SPAT.

Anyone know of a multi-channel AU-host that might do the job? Thanks in advance.

_

[Lawrence]

Reaper?

[tack]

Not just that. But bringing up to eight sources into one single SPAT-space and having these sources, all of them together, influence the room response for the whole group. That's fundamentally different from giving each source its own SPAT treatment.
A room responds differently to a bass than it does to a flute. It'll also respond differently to the combination bass+flute. And SPAT can do that with up to 8 sources.

Are you sure it works this way? Our intuition about the physical world suggests it should be so but there's every possibilty that given the reverb function f, f(a+b) = f(a) + f(b). I tried to determine this for EAReverb2 but it doesn't pass a null test which would be a prerequisite to do a black box test for this behavior (clearly there is some randomness in the reverb function). I didn't try this with SPAT and my demo has now expired, but if it passes a null test we could learn whether multiple sources do interact within the same instance. I'd be very interested to know.

It's unfortunate you would have to go through such rigmarole to send multiple sources to SPAT. It seems like it should be a basic feature of every DAW. (REAPER had no trouble with it.)

[Piet De Ridder]

Jason, I would think there'd be a substantial difference between a room responding to 8 different instruments simultaneously, or that same responding to all these instruments separately and then summing the results. No?
And since SPAT is not simply something which you add to a source (like you would a traditional reverb), but actually emulates the phenomenon which occurs when a source generates sounds in a space, I'm inclined to think that there is also going to be a big difference between sending 8 different sources into SPAT all at the same time, on the one hand, or processing them all individually, on the other.

Larry, Reaper you say? Never looked at the app, don't know anything about it, but that's about to change.

Thanks both!

_

[tack]

Jason, I would think there'd be a substantial difference between a room responding to 8 different instruments simultaneously, or that same responding to all these instruments separately and then summing the results. No?
And since SPAT is not simply something which you add to a source (like you would a traditional reverb), but actually emulates the phenomenon which occurs when a source generates sounds in a space, I'm inclined to think that there is also going to be a big difference between sending 8 different sources into SPAT all at the same time, on the one hand, or processing them all individually, on the other.

I do agree that in the real world separate sound sources would have some interaction and would be more than the sum of their parts. Is SPAT implemented this way, such that the reverb function is distributive? I don't know, but it isn't necessarily the case. It could be tested if SPAT passes the null test (i.e. two separate identically configured instances with the same input would need to generate the same output). Barring that, or someone from Ircam confirming the implementation, a blinded listening test perhaps.

[Piet De Ridder]

(...) Is SPAT implemented this way (...)?

Yes, it is. In fact, it is even more complex than that, in that SPAT's response generation (reflections & reverberation) doesn't just depend on the type of source, but also on its position in the space which SPAT is set to simulate, and also on the width and angle of the source's sound projection. In other words: a tuba in the centre with its projection towards the listener will make SPAT generate a different room response then when that same tuba is, say, positioned somewhere to the side and is oriented in some other direction. That is logical, no?

All of which suggests to me that there should be a significant difference between, on the one hand, SPAT processing a group of (positioned) sources, and, on the other, SPAT processing individual sounds (even if these are, individually, positioned and rotated in exactly the same way as they are in the group).

And by 'significant' I mean: certainly MUCH more noticeable than a difference which only a null-test can reveal, because if it were no bigger than that, I'm not in the least interested in pursuing this whole thing anyway. I'm not going to complicate my setup if the gains can only be demonstrated by a null-test. Anyway, I'll keep you posted, after I've spent some time in Reaper.

_

[tack]

All of which suggests to me that there should be a significant difference between, on the one hand, SPAT processing a group of (positioned) sources, and, on the other, SPAT processing individual sounds (even if these are, individually, positioned and rotated in exactly the same way as they are in the group).

This is what I mean by f(a+b) = f(a) + f(b). Given a reverb function f and two input sources a and b, do both sources positioned independently in the same SPAT instance, i.e. f(a+b), differ from two separate instances with the same configuration, i.e. f(a) + f(b)? We're hoping they're not equal, because that would better match reality where sounds within a room will interact with one another.

But I do disagree if you're suggesting that because SPAT has sophisticated room simulation for an individual source it follows that it will also simulation interactions between sources in the same instance. It'd be cool if it does, and maybe SPAT, among other plugins of its ilk, is the most likely to have implemented it.

And by 'significant' I mean: certainly MUCH more noticeable than a difference which only a null-test can reveal, because if it were no bigger than that, I'm not in the least interested in pursuing this whole thing anyway. I'm not going to complicate my setup if the gains can only be demonstrated by a null-test. Anyway, I'll keep you posted, after I've spent some time in Reaper.

I wish my demo was still functional. I could easily put some samples together for a blind test. Unfortunately I didn't think to test this until after my demo expired. I look forward to your results!

Even if it ended up being acoustically identical to separate instances (and so f(a+b) = f(a) + f(b)), another benefit to using multiple input sources is lower CPU usage thanks to fewer instances. When I was playing with the demo, I did confirm this was appreciable. (I tried 4 sources.) Although if it's as cumbersome to setup with your DAW as it sounds, the CPU difference would probably not be worth that much added complexity.

It sounds like you're going to try routing instruments out of your DAW, into REAPER running SPAT, and then output back out into your DAW? Even if that works it still sounds terribly complex.

[kpc]

.

Anyone know of a multi-channel AU-host that might do the job? Thanks in advance.

_

How about Ozone? I don't have it, but can't it run AUs as stand alone? I don't know if it's multichannel.

Or Bidule?

Cheers
kc

[Piet De Ridder]

Kayle, thanks very much. But my first port of call is getting to know Reaper well enough to be able to do a simple test with a few instruments. If the results are musically meaningful (the difference between one SPAT-instance per instrument, and one SPAT-instance for the group, I mean), I’ll dig deeper into finding the most suitable host. If the difference is negligible, I’ll keep doing what I’ve been doing so far.

Jason, I’m pretty confident there will be some difference for the simple reason that f(a+b) results in one room response (of two sources), whereas f(a)+f(b) gives two responses (of one source each) which, inevitably, are going to be stacked/summed in the mix. I’m no mathematician, but surely, that has got to sound different? Anyway, as I said, I’m going to do some testing in Reaper first.

_

[tack]

Jason, I’m pretty confident there will be some difference for the simple reason that f(a+b) results in one room response (of two sources), whereas f(a)+f(b) gives two responses (of one source each) which, inevitably, are going to be stacked/summed in the mix. I’m no mathematician, but surely, that has got to sound different?

It's not a certainty, though. It depends entirely on the distributive property of f.

If SPAT's algorithms don't simulate any interaction between sources a and b, then indeed SPAT would only just sum the results same as the DAW's mixer.

I don't have a lot of reverb plugins, but I've tested this with ValhallaRoom, ValhallaPlate, EAReverb2, and Reaverberate (REAPER's built-in reverb). Once I establish it can pass the null test on a single source, I send two sources each to their own separate reverb instances, and then again together to another shared instance with phase inverted.

• ValhallaRoom: requires late mod depth to 0% to pass the null test. Shared instance is the same as two separate instances.
• ValhallaPlate: same as Valhallaroom
• EAReverb2: requires late reverb disabled to pass null test. With only ERs, shared instance is the same as two separate instances.
• Reaverberate: shared instance is the same as two separate instances.

So, for every reverb plugin I own, it's indeed the case that f(a+b) = f(a) + f(b).

Now it's difficult to compare the above to SPAT, because SPAT obviously receives the sources independently, whereas in the above tests, the sources are summed by means of the send before they reach the plugin. So SPAT has a real shot at modeling this properly. But it does at least show that it's up to the implementation and can't be concluded a priori that it will work according to our intuitions.

[Aesthete]

installed and working pretty good here just as a basic replacement for Soundflower.

I'm going to run my itunes library through some cool tube plugins : P)