Norns: mlr



Random bug on my device with latest firmware. I had mlr open for about a week with five or six buffers and a grid attached. I came home and the OLED was turned off and the grid was no longer responsive. All keys on the device were unresponsive. The power LED on the back was orange. I plugged in the wifi dongle and the rear LED blinked white a few times so the device was still on.

I don’t have a serial debugging setup so I couldn’t get any information from the device. Doing a hard reset made everything better.


Maybe this is already possible or just a slight modification in Lua, but I find myself wanting LP and HP filters per track - are there already plans for this?


serial login in very simple, no debugging environment needed. from the docs:

other access

When connected via WIFI you can SSH into norns at the IP address shown in SYSTEM.

  • user: we
  • password: sleep

Without WIFI, you can connect to norns via USB-UART by connecting the power cable to your computer. On Mac/linux do:

screen /dev/tty.usb(tab) 115200

Where (tab) appears hit TAB to autocomplete the serial number. Login is the same as above.


That’s my wish as well.


yes there are (in softcut rewrite) - no its not a simple scripting change


This is insanely useful. Apologies for being bad at reading comprehension yet again.


curious as to what the issue was that prompted this change:

only had norns for a couple of weeks, so I’ve only known this version of mlr with 4 tracks



[ ]

nutshell: supernova on ARM causes hard crashes with too-high CPU usage. reducing voice count was a stopgap while we build routing, softcut and mix effects from scratch. (PR #650, 661.)

supercollider will continue to be supported for user engines (as will other environments like puredata) but won’t be a system dependency.

in current WIP scheme, softcut and mixer/fx are separate jack clients and can be run on separate CPU cores, without the complex DSP worker-thread system introduced by supernova. (additionally: new softcut implementation corrects a lot of things and can do e.g. rate modulation while recording, &c; the setup also allows softcut to process engine output. )


I can help test some of this if I know where to start. Read through the PRs on GitHub so I kinda get the gist.


Wow this is super cool. Not the supercollider limitations, the impressive work on “native” softcut.


Has anybody figured out how to max rec time per track or otherwise to slowdown mlr “sequencer” ? Thanks


Im running the latest firmware 181101 and loading in 4 x 10 sec wav clips into 4 slots on 4 channels.
They play fine.
I save this as a preset but on recall the 4 wavs do not load as part of a preset.
Any ideas if the wav recall was supported as i want to make a liveset?


I think there would need to be something to set/save the current wav file to a parameter and then that would save along with other params.

so something like params:set(y..": sample", path) (y being the channel or row)

and then it would probably also need to be sure to load those on script init.


I think the “MLR clip saving” feature that was added in update 181101 is a separate feature from “recall” I think recall is more for quickly changing configurations during performance. Haven’t figured out clip saving though. The feature I am most looking forward too is the ability to launch groups of clips so that I can perform premade songs while leaving a few tracks for live sampling


Yes, clip saving and recall are separate functions and not what I mean’t. I want t save and load a session with prerecorded audio samples so I don’t have to manually load in 16 samples everytime I switch on mlr, like you can already do on Foulplay and Glut.


indeed, i’m working on this feature (pset saving with file loading) to be ready along with mute groups. thanks for your patience everyone.


any chance the new MLR will have an input selection for what input channel writes to which buffer and output selection for panning?


Yes to both of those 20ch


Can’t wait :slight_smile:


Oh I guess I should clarify that I’m talking about engine features. I won’t work on the MLR script and shouldn’t make promises about what it exposes as parameters. Those two features were in fact already available in the “old” engine. However I have made voice panning explicit (level, pan with equal-power law) instead of (level, level)