What are you working on?

[vurt]

Thought i’d start a relaxed general thread where you can share what you are working on. Ideas, issues, whatever you feel like sharing.
Maybe someone is doing the same port as you are thinking about doing?

I use AI to code.
So far i have made:

Trigonomicon - a dream module of mine, no clue why no one has made it, it’s what got me into doing these modules. It’s a drum pattern generator with 60 (so far) built in patterns for jungle, breakcore, DnB, Drillcore, IDM etc. 32 steps. It has a mutate function for endless variations of these patterns. It’s awesome and i love it.

Slidewyrm - ported version of the really wonderful TB3PO (from ornament & crime) with some additions: accent out, auto accent on either step 4 or 8 or random. Slide amount, Gate length - both CV:able.

Septagon - a 7/4 drum generator (no built in patterns, entirely generative). Struggling slightly with performance here, the other two has great performance, not this one. 18-20% at 48KHz, so not good at all i think, will give it another pass. Has some cool features like generated velocity and outs for this. I do like the Trigonomicon better, so far.

Upcoming (maybe)
Drum multi-sample player - should have trigger sensitivity, maybe each drum has 3-4 samples and they will sound different depending on voltage sent to trigger. Not sure there exists one already or if someone is working on a similar thing?

IDUM (Mystic Circuits)
PreenFM2 (this is my favorite FM module, i use it for all my tracks)

No clue if i can make them, but i am going to try. PreenFM2 might be hard, but the AI who looked at the source code said it was doable; rated it “easy-medium”, IDUM it rated “easy”, so hopefully at least IDUM can be ported.

2x Bemushroomed modules

[vurt]

I have a feeling this is going to be my “blog” lol.

Ported PreenFM2. Much harder than e.g TB3PO, took a few hours. It was the MM port which was difficult, translated very easy to VCV i would say.

The original hardware version (in eurorack) has issues with glide, i made the AI take a look at that and it was basically down to the MIDI-implementation which did not translate over well to voltage controlled. It fixed this, which is great since i really miss having glide.

However, there is another issue with glide (also on the euro hardware) that i must look into, its down to use of glide and browsing patches while the unit is playing a voice, this can “crash” the voice and you have to restart the unit.

This version is stripped of Arp, filters, editor, midi. I thought it was probably a good idea to make it as minimal as possible. It can load my .bnk files fine.

So far it has 4 CV ins to shape the sound. you can also use a knob to switch Algo (the operator selection). So even if stuck to my banks you can still shape the sound somewhat. I might expand upon it too, any ideas are welcome.

Still early, might discover bugs, so far from finished.

Here is a demo from VCV Rack: PreenFM2 ported to VCV Rack and 4ms Metamodule - YouTube

[Ross]

Outside of the 4ms library, for Moffenzeef stuff I’ve been playing around with a bunch of concepts based on preset organ drum machines. Much like your trigonomicon, I find that people actually love when other people help make music for them, so I support more modules like this, especially for Meta!

Do you have any interest in porting DitDit for VCV/Meta? I have absolutely no interest in doing it, as I’m spinning 300 other plates at the moment, but think it would be nice addition for both VCV + Meta. I already made a branch on my VCV repo with the UI for the DitDit, and the code for the hardware version of the module is also available online. If you’re interested, I can send you a free hardware one in exchange for the job. The keyboard feature set is pretty dense and it requires a custom screen / animation which is why I’ve been putting it off. I imagine it will be an intense undertaking. It would be a fun job for a Meta dev fanatic

Let me know!

[vurt]

That’s a neat idea! Haha, those are really cheesy in a fun way!

I have so many of my own ideas right now + also entirely out of AI tokens until next month, i have spent way too much in 3 days (yes i can’t program, i need my AI babysitter, and they cost money, sigh!) So I will unfortunately have to say no to that, sorry! Sounds like a cool module though!

[vurt]

Against my better judgement (just purchased the MM audio expander) I purchased even more tokens. Whatever, i am having fun and i had some ideas i couldn’t wait to try out for MINIMALITH.
Who needs food anyways.

Since we are in VCV Rack, or on MM, we don’t want advanced deep digital parameter editing, so instead i added a MORPH function, and a save over currently loaded preset.

MORPH will morph 2 random presets from the currently loaded bank. The results can be really amazing and useful, i have come up with some of my favorite patches in the last hour just using that.

If that’s not enough i added (trimpots and CV-in)

ALGO
Switches the Operators.

EVO (Evolution)
introduces controlled variation to the patch so repeated notes don’t sound identical. As you increase EVO, the synth gently “evolves” aspects of the sound over time and/or per note, making the timbre feel more alive and less static.

IM SCAN (Index / Modulation Scan)
shifts the character of FM by scanning through modulation intensity relationships. In practice, it changes how “hard” operators modulate each other (the FM “index” behavior), which can move a sound from smooth and tonal to bright, metallic, or noisy.

DRIFT
Adds pitch and phase instability similar to analog oscillator “wander” and voice-to-voice detune. It’s meant to thicken the sound and reduce perfect digital precision.

design is not finished.

[Submit]

Nice sounds!

[TheTechnobear]

@vurt

Against my better judgement (just purchased the MM audio expander) I purchased even more tokens.

a suggestion to help, you may be already doing this - but it may be useful for others that may be starting out with llms for coding

what LLM are you using?
as I mentioned previously, I’d highly recommend github copilot with vscode.
one reason, this is a good combo, is you can freely switch between different LLMs including a few that are free (grok fast, gpt-4.1 , gpt-5 mini etc ).
these are not as good as paid offerings, however, you can use them for more ‘mundane tasks’.
I found this could dramatically reduce token usage.

e.g. in your case, use the paid versions (e.g. claude or whatever you prefer ) to do the first generation, then use the free ones to iterate/ refine. you can always switch back, if you are not getting what you need.

practical note: vscode by default has ‘auto’ for LLM selection, a simple dropdown - this does give 10% discount, but its so easy to move to a free llm where needed. remember switching llm will drop context.

also as above, prompting (and context) is everything, if you are inefficient you will burn thru a lot more tokens. one detailed prompt getting you 90% there, is much cheaper than 3 smaller prompts where you iterate to the same point.

I learnt this the hard way , when undertaking a really large refactoring (boring work for a developer!) project. I was burning thru tokens at an alarming rate, to the point I though I was going to have to stop, but then I changed my workflow, and started selectively using free llm, and tokens started lasting days on end.

I’ll admit, the downside is, its not quite as frictionless as using auto or your preferred (paid) llm, and for small jobs, its tempting to just say ‘what the hell’.
but its an option to make the tokens last longer

reminds me, talking of ineffecient, I need to get back on this… as Ive got a monthly plan Ive not used for about 3 months

[vurt]

@TheTechnobear

github copilot with vscode.

Indeed, that’s the best one. Before this i took a Cursor subscription, burned through all the tokens in one day and i had burned all the Copilot ones in 2 days, lol. But yes, a lot is down to not wasting tokens on models which are 3x the cost, and i did do that for the PreenFM2 port, probably not worth doing but i wanted to be sure it was getting the best treatment possible

I do not recommend Cursor actually. It has a lot of hype, but it’s super, super expensive in usage. I purchased 10 euro extra now with Copilot and even after many, many hours i have only used around 2 euro. Though i do use Chatgpt 5.2 now and not some 3x model.

[vurt]

New module: AMENOLITH

Presenting AMENOLITH for VCV Rack and 4ms Metamodule

AMENOLITH, a sample-player (bundled samples).
Each kit has 3 LAYERS for each drum. Two different ROUND ROBIN functions to change which of the 3 drums will play (basically low-medium-strong hit), or Velocity CV-input.
HUMANIZE function.
DRIVE to overdrive the sound.
LENGTH, PAN, PITCH for each drum.

[vurt]

A new Phase-mod synth with wavetables.

Edit, new demo (now with user wavetables)

[kuttor]

So I develop on the Disting NT and the Nebulae v2. And bought a Metamodule to port my stuff over. kuttor (Andrew) · GitHub. I do mostly physical modeling drum stuff right now. that are highly influenced from WMD drum modules but with a very high focus towards physical modeling… for instance El Applauso is a hand clap physical modeler using HelmHoltz resonator and a granuler synth engine to create multiple hand claps… what I call my CrowdEngine. There is also a venue simulation (reverb but with room modeling) so studio is highly treated… basement is highly reflected, arena is airy.

[vurt]

How cool! Any sound examples of the hand clap?

I wish someone would do a super realistic snare drum model. Something like the Nord Drum 2.

[kuttor]

Dude. Totally can give ya examples.

Matter of fact. Would love to get with someone to figure out how to port these.

And yeah. I been working on a snare. And a cymbal one which is kicking my ass called Los Plato’s (the dishes)

[vurt]

I’m a hack/fraud so i port with AI.. Might want to do it more properly, idk.

Would love to hear examples of both of these, sounds super interesting.. I’ve owned quite a few of the WMD ones, Crucible etc, ultimately i sold them and sticked with samples, sometimes i do regret selling them.

[vurt]

a ton of new things was added for Phaseon, getting closer to something pretty decent

Phaseon for VCV Rack and 4ms MetaModule - YouTube

[vurt]

A New module:
Xenostasis - mix of Byte Beat, FM, Wavetable to create either drums (if using the trigger in) or alien textures / drones. Can actually sound kind of melodic too with some of the Byte Beat stuff.

Phaseon and Trigonomicon also got the faceplate updated.

All ported to MM. Xenostasis ~30% CPU at 48KHz, quite a bit higher than the others (around 15-20%)

Edit:
Long demo

[guitooo]

Hi @vurt
thanks for your blog lol
A quick shout to say that I’m impressed that you create your own module with AI.
Your modules sounds great, how can we try them out on Meta? (excuse my innocence )

[vurt]

Glad you enjoy!

Well i finished all the faceplate graphics, so what’s left now is;

I need to make 4 more banks for Amenolith
1 solid bank for Minialith
Make readme’s and manuals.
Test and tweak more, make sure the Metamodule variants works and are as performance friendly as they can get.
Make sure i have understood to 100% the licensing for TB3PO and PreenFM2 ports
Put it up on Github

It certainly won’t release this week, maybe in 1-3 weeks i reckon. I’ll try to not make any more modules now, just hyper focus on these so i can release them.

[vurt]

The full line-up with close-to finished faceplates

[guitooo]

Thanks for your reply

Solid roadmap you’ve set

The design is really nice too
I’ll follow up for sure!

Good luck for the final rounds