You want the flexibility of PC software, but the performance of standalone gear? A new music OS is the latest effort to promise the best of both worlds.

Sure, analog gear is enjoying a happy renaissance – and that’s great. But a lot of the experimentation with sound production occurs with software (iOS or Windows or Mac) simply because it’s easier (and cheaper) to try things out on an Intel or ARM chip. (ARM is the architecture found in your iPhone or iPad or Android phone, among others; Intel you know.) Some manufacturers are already making the move to standalone hardware based on these architectures – at AES last year, I saw Eventide’s massive coming flagship, which is totally ARM-based. But they’re typically rolling their own operating system, which provides some serious expertise.

MIND Music Labs this month unveiled what they called ELK – a Linux-based operating system they say is optimized for musical applications and high performance.

That means they’re boldly going where… a lot of players have tried to go before. But this time, it’s different – really. First, there’s more demand on the developer side, as more makers have grown intrigued by off-the-shelf CPUs. And developer tools for these options are better than they’ve been. And hardware is cheaper, lower-power, and more accessible than ever, particularly as mobile devices have driven massive scale. (The whole world, sadly, may not really feel it needs an effects processor or guitar pedal, but a whole lot of the world now has smartphones.)

ELK promises insanely low latencies, so that you can add digital effects without delaying the returning signal (which for anything other than a huge reverb is an important factor). And there are other benefits, too, that make music gadgets made with the OS more connected to the world. According to the developers, you get:

Ultra-low latency (1ms round-trip)
Linux-based, using single Intel & ARM CPUs
Support for JUCE and VsT 2.x and 3.x plugins
Natively connected (USB, WiFi, BT, 4G)

That connectivity opens up possibilities like sharing music, grabbing updates and new sounds, and connecting to wireless instruments like the ROLI line. There’s full MIDI support, too, though – and, well, lots of other things you can do with Linux.

(JUCE is a popular framework for developing cross platforms, meaning you could make one really awesome granular synth and then run it on desktop, mobile, and this platform easily.)

Now, having done this for a while, I’ve seen a lot of claims like this come and go. But at least ELK last week was demonstrated with some actual gear as partners – DVMark, MarkBass, and Overloud (TH-U).

1ms latency claims don’t just involve the OS. Here, ELK delivers a complete hardware platform, so that’s the actual performance including their (high-quality, they say) audio converters and chip. That’s what stops you from just grabbing something like a Raspberry Pi and turning it into a great guitar pedal – you’re constrained by the audio fidelity and real-time performance of the chipset, whether the USB connection or onboard audio. Here, that promises to be solved for you out of the box.

DVMark’s “Smart Multiamp” was the first real product to show off the platform. Plugin Alliance and Brainworx have signed on, too, so don’t be surprised if you’re soon looking at a dedicated box that can replace your laptop – but also run all your plug-ins.

And that’s the larger vision here – eventually ELK has its own plug-in format, and you should be able to move your favorite plug-ins around to connected devices, and access those gadgets from Android and iOS, But unlike using a computer or iPad on its own, you don’t have to sweat software upgrades or poor audio performance or try to imagine a laptop or tablet is a good music interface live.

This leaves of course lots of questions about how they’ll realize this vision and more questions if you’re an interested developer or manufacturer. I’m hopeful that they take the Eurorack market as a model – or even look at independent plug-in and app developers – and embrace a model that supports imaginative one-person developers, too. (A whole lot of the best music software and module ideas alike have come from one- and two-person shops.)

I at least like their vision – and I’m sure they won’t be alone. Best line: “Whether your idea of music is to be shut in a studio that looks like the bridge of a Klingon cruiser or you are a minimalist that wants everything to sound exactly like in 1958, we think you will be surprised at just how much smartness is going to affect us as musicians.”

I’ll throw this out here for now and let you ask away, and then we can do a follow-up soon. Loads more info at their site:

https://www.mindmusiclabs.com/elk/

  • 1 ms latency sounds like a dream!

  • itchy

    a more tighter software system gear for audio and video sounds like a nice idea there is a lot of bloatware and bullshit running on our comp and taking up resources . sounds like a solution. dear ableton keep an eye on this guy

    • PaulDavisTheFirst

      people have been building systems with these sorts of capabilities, enabled by linux, for many years. the particular combo/format that Elk is doing is somewhat new, but you can be pretty certain that computers running Linux and processing audio can be found inside most major audio software companies these days.

      • Apoclypse

        Including the MPC Live which does this already. ELK is nothing new in-terms of functionality. What’s new (I think) is that it’s essentially an off the shelf OS built specifically for people like Akai or any other vendor who may not have the expertise, time or budget to create their own custom OS or Linux solution. They license ELK and build their tools on top of it.

  • This is awesome. I’m currently working on a looper using chuck and a raspberry pi. Even though most of the latency can be compensated for (by looking ahead for the initial loop and behind for audio latency playing to existing loops), I have to use hardware monitoring through the soundcard for the “realtime” playing. Even still, jack audio on linux will randomly act up and give glitches no matter how much I tweak, disable and reduce services. Really looking forward to seeing where this goes!

    • Michael Heuer

      Hello Jordan! I’m also interested in ChucK + Raspberry PI, curious to see what you’ve been able to come up with. Ping me on github at https://github.com/heuermh/lick

  • Elekb

    Awesome stuff.

    I wonder about a few things:

    – is the 1ms latency for real? Has this been tested thoroughly? Is that feature on warranty?

    – Can ELK be connected to a screen, keyboard, MIDI peripherals, hard drives etc. and just work as a regular computer when you’re at home?

    – Will this replace PCs for good? (hint: probably not just yet, desktop PC’s still have way more horsepower).

    – Also, will this finally be the end of guitarists and keyboardists with unreliable and unwieldy iPad setups? Hope so.

    – Finally: where / when can we order one of these beasts?

    • Polite Society

      From the website to answer your second question: “We have a generic framework with a component library for many kind of controls: buttons, switches, analog potentiometers, rotary encoders, keyboards, etc. It is possible to customize the behaviour by simply editing a JSON file or with a real-time IPC system, saving lots of development time for your product. Alternatively, you can use other ICs of your choice and communicate to ELK with UART, I2C or SPI.”

    • Stefano Zambon

      Hi, 1ms latency is analog-to-analog, measured properly (using a PC running jack_delay which has 1 microsec measurement accuracy), and with hard real-time guarantees (i.e. no jitter no drop outs if the plugins code stays below max allowed CPU usage). At NAMM we were running the measurement live while the system was running at our booth.

      ELK is not available as a customer product, at the moment we’re discussing with HW/SW companies to license the solution for their products.

      Cheers,
      Stefano (MIND Music Labs)

  • PaulDavisTheFirst

    you’ve been able to do more than this on Linux for years (finding the right hardware was difficult, but …).

    and yet .. what is is that actually stops people from changing platforms?

    “can i run ?” – NO
    “can i run ?” NO

    it’s really great to see some people taking the low latency capabilities of Linux and trying to package them in a way that makes them potentially more usable and accessible.

    but for more than a decade we’ve see INCREDIBLE resistance on the part of developers and users to switching to a linux-based platform, and i’m skeptical that what Elk is doing here will change that.

    btw. there’s another thing to compare this: the Mod. Elk is like a Mod with a touchscreen 🙂 Both are carefully constructed hardware running a carefully constructed Linux kernel; both are intended to run plugins as much as anything else.

    • Polite Society

      Because it’s tightly linked to JUCE, you can develop on any platform you like (I assume). You’d just be running it on a ELK as a destination.

      • PaulDavisTheFirst

        That’s already true for Linux in general. Do you see a vast flood of developers choosing to release their JUCE-based tools for Linux (and there are a LOT more Linux users than there will be Elk users for some time) ?

        • prrrrt

          One reason for that is probably wine, some stuff just works with a vst bridge and I’ve been given that as a reply from some devs before. Since then, I’ve seen some people switch since Bitwig was available on Linux, as they wanted an Ableton live style DAW, which brings its own set of “devices”. Other than that, true. You can see plenty of juce plugins, where the devs just went “nope”. Hm, if AVB devices work, maybe I should replace my aging traveler plus ultralite mk1 setup, which when chained never worked a 100% reliably on win or Linux.

          • PaulDavisTheFirst

            two clarifications regarding the ultralite:

            1) I don’t use it in AVB mode, just as a class compliant USB interface
            2) there’s a bug in the the driver and/or device in which the input channels get randomly remapped (in blocks of 8). So channel 8 can end up as channel 1, or channel 16, and sometimes as channel 8.

    • Malg Amoe

      True, for me the biggest problem is to get really good audio interface that runs on linux, with also the gui to change all the parameters of it.
      Like a rme, antelope or the like.
      I like the idea of that platform though.

      • PaulDavisTheFirst

        I wrote drivers for the RME Hammerfall series back in 1999/2000 (Digiface. Multiface etc). They’ve worked ever since. RME firewire devices never used standard protocols, so were not well supported; ironically now they are, but they’re also out of production.

        All USB class compliant devices just work on Linux (why? because the iPad’s “no drivers” policy has forced manufacturers to actually comply, which didn’t happen before). Currently I have a MOTU AVB unit plugged into my main work system. Current RME USB devices will work when in class compliant mode (I believe this is true of Antelope as well).

        The number of people who just plug in newer USB interfaces to a Linux system and find that “it just works” grows every day.

        • Malg Amoe

          Yes it’s true, but when you have crazy routing available and dsp to control with some editor and you can’t. It kinda suck. But I’m sure we’re getting there.

          • PaulDavisTheFirst

            That’s why moves like MOTU (of all companies (*)) to use a web interface for device configuration is important and encouraging.

            The MOTU Ultralite configuration system is amazing, and this design approach should really be taken up by every audio interface maker. Configure the damn thing from your phone, your tablet, your laptop, your desktop, some NSA supercomputer cluster. All it needs is a browser and a network connection.

            No more “Device Configuation Utilty” applications that have to be written for multiple platforms (and hence no more “you can’t do that from Linux”).

            (*) in the early 2000’s MOTU were particularly personally insulting to me and other Linux audio developers, and for 2 decades displayed absolutely zero interest in getting their devices supported on the platform.

          • Malg Amoe

            God you’re so right and also it ensure you will be able to run it in 20 years without problems. I really wish they started move to this. How are the preamps on the new motu, i have the ultralite mk3 and those suck. Having used the zen studio of a friend for 2 years in studio, i absolutely fell in love with the sound this thing has. So I am really tented by the new discrete serie of antelope and just whish i could program the dam thing with anything i have around.

          • PaulDavisTheFirst

            I’m the wrong guy to ask. I just develop a DAW (Ardour (and Mixbus and Waves Tracks Live and …), so my concerns about audio quality are small to insignificant. I just need channels and the ability to do basic work in my own small studio.

          • Malg Amoe

            Mixbus is awesome, so thanks for making it possible!!

  • Polite Society

    Interested in cost of development kits/licenses. I see that’s they say it’s possible to run it on any processor that can run linux from the last 5 years, but would be interested in a kit that had a handful of inputs/outputs/kobs/etc and maybe with a small display or video out to use for prototyping.

  • mavin

    PM 1998 and BeOS 😉

  • Diez Roggisch

    It’s maybe worth pointing out that the VST support is *not* capable of just taking any beloved vintage 32 bit Windows plugin (or macOS for that matter) and run it. It’s “just” capable of running Linux-built plugins that conform to the VST standard. So this needs developer adoption first.

    Also I doubt the claims about them being able to do something significantly different regarding ARM-audio than the rest of the world. ARM SoCs usually come equipped with I2S-busses to interface with audio circuits. This is what the PI uses, and this is 100% what they use. So while the significant work of designing and tuning a system quite a feat, it’s the same as people have done with the PI or other ARM-based boards.

    The offering is of course still cool as it relieves a lot of people from the burden of aforementioned design and tuning, and allows to focus on their user value in form of effects/synths.

  • chaircrusher

    what is missing in the Linux world is something that costs money: productizing. In other words, exhaustive testing, professional user interface design. If these guys can do this, good on them.
    But this smacks of hype. Linux is not a real time OS; making it into one is a well known edifficult problem. I haven’t read theiraus marketing materials but I bet they stop short of guaranteeing any particular latency, because Linux doesn’t work that way.

    There are low latency kernel mods, have been for 20 years. But guaranteed latency is a myth. if you want that you’ll get there faster by abandoning Linux than by patching it into submission.

    • PaulDavisTheFirst

      Linux is entirely capable of being a real-time kernel. Almost all of the RT patches are in mainstream kernel now. The current RT patch is quite small, and most major distributions offer the RT kernel as an alternative kernel.

      It is true that Linux is not a smart choice for hard realtime systems (where failure to meet deadlines has substantive implications). But for audio, Linux has been able to meet realtime deadlines better than MacOS (nee OS X) or Windows for more than 15 years now.

      I’m not sure you know enough about realtime and Linux to make these comments.

      • Diez Roggisch

        And there is XENOMAI if you can get it to run on your concrete platform. This allows Bela.io to also reach the 1ms roundtrip. PREEMPT_RT should be good enough for less extreme use-cases though.

        • PaulDavisTheFirst

          Xenomai isn’t quite what I would typically mean if I said “real time linux”, and my understanding is that device I/O requires quite some work (i.e. if you have a particular PCI card you want to use, it’s not necessarily true that the ALSA driver for it will be of any use). On the other hand, if you use a driver that Xenomai already has “converted” then sure, this is a really solid and quite nice solution. It would be lovely to see more platforms based on bela.io appearing …

        • Stefano Zambon

          Hi, we have similar solutions and performances on ELK.

          The main differences compared to Bela is that we support a wider set of CPUs which provide enough power for serious audio processing, plus VST hosting, a headless DAW and other stuff.

          Stefano (MIND Music Labs)

  • R__W

    Someone else mentioned this but it’s not really going to work. You won’t be able to just drop in a VST. The VST developer will have to provide a version of the VST that works with this system. Nobody is going to do this.

    • PaulDavisTheFirst

      Well, “nobody” is not quite true.

      http://linux-sound.org/linux-vst-plugins.html

      But any VST plugin that you might care to name (and run)? Yeah, quite probably not. There’s also the non-small detail that the system is ARM based, whereas most (all?) of the plugins in the list above are released for x86/x86_64 systems.

      • Stefano Zambon

        Hi Paul,
        just a clarification: current ELK dev boards and the DV Mark product are based on x86 Atom CPUs, not ARM.

        VST plugins need to be recompiled to run on ELK, but if written properly (i.e. no nasty things on the real-time thread), the process is easy if the plugins runs already on Linux. For instance, at NAMM we were demoing our dev boards running a bunch of open source plugins (MDA, Dexter, etc.) which were recompiled without changing their source code.

        Disclaimer: I am one of ELK’s developers, feel free to ask any questions.

        Stefano

  • Axel

    1ms latency on an embedded arm platform has been here for a year at least: bela.io
    I have one but I haven’t gotten around to really dive into it yet. Imagine what could have happened if these guys would have joined forces instead of competing in an already miniscule market.

  • Unless it runs Max and Reaktor and [insert DAW here] it won’t replace a laptop. Otherwise Linux would already have stormed the studio. And unless it’s cheap as chips it won’t be replacing Raspberry Pi and other microsystems.

    It’s apparent the developers know this, and are promoting ELK as a platform for embedded systems. It will be hidden away under the surface of hardware tools. Musicians are unlikely to ever know ELK is in their studio.

  • Miles∙Prower

    I know I’m jumping into this discussion late in the game, but I wanted to give a shoutout to QNX, a real-time UNIX-like OS that powered a number of audio products back in the day — and is still in use for real-time applications across various industries. My venerable Sony DMX-R100 console runs on QNX, and even though the console has far less computing power than any current smartphone, the console offers a ton of functionality at very low latencies, and it can access its file-system and do other non-audio tasks without a hiccup in its real-time audio processing.

    https://en.wikipedia.org/wiki/QNX