October 28

Vibe Coding for MIDI Hardware: The Future of H12MIDI Pro

28  comments

What If You Could Use Vibe Coding to Customize Your MIDI Hardware?

The CME H12MIDI Pro puts powerful MIDI connectivity and customization in your hands. But what if the next step was creating your own MIDI features using AI?

Meet the H12MIDI Pro

The CME H12MIDI Pro is designed to be the central hub of your MIDI setup. With six USB host ports, six pairs of TRS MIDI inputs and outputs, and USB-C connectivity to your computer or mobile device, it brings your MIDI gear together in one powerful standalone device.

Using the free HxMIDI Tools software, you can customize MIDI routing, filtering, merging, splitting, and mapping to build a setup that works exactly the way you want.

No complicated workarounds. Just flexible MIDI control.

What If You Could Go One Step Further?

We're entering the era of vibe coding, where people can use AI to create software by simply describing what they want.

Now imagine applying that idea to MIDI hardware.

What if you could tell an AI assistant:

  • "Create a custom MIDI function that switches between my synth setups"
  • "Build a MIDI processor that transforms incoming messages into something completely different"
  • "Make a custom controller function for my live performance"

And what if you could share your creations with other musicians, or download features created by the community?

Your MIDI hardware could become more than a device with predefined features. It could become a platform for creativity.

How Open Should MIDI Hardware Be?

Of course, there's an important question.

Should manufacturers allow users to customize parts of their firmware, or should certain functions remain protected to guarantee stability and reliable performance?

Perhaps the future lies somewhere in between: a stable hardware foundation with a flexible layer for community-created features.

We Want Your Ideas!

Using the H12MIDI Pro as an example, we'd love to hear what you think.

If you could create your own MIDI features using simple AI prompts, what would you build?

Would you want complete freedom to modify firmware? Or would you prefer a protected system that allows custom functions without compromising reliability?

Join the conversation, share your ideas and concerns in the moderated comment section below.

The future of MIDI hardware is created together.


Tags

crowd creation


You may also like

Getting the Music Flowing Again

Getting the Music Flowing Again
  • I honestly would prefer if no AI was involved in the creation of music and its gear. AI in music has more often than not been used by stealing from artists that have no say in whether they want that or not. Making music generative is cool but not when AI is involved. If AI gets used in gear from CME I am boycotting them

  • Hi, I’m Matthew Whitaker. I’m a blind musician, I play keys, drums and a lot more, and I build accessible tools for my own studio.

    This gets me excited mostly because of accessibility. A lot of MIDI setup happens in visual editors with drag-and-drop routing, and those can be really hard with a screen reader. If I could just say what I want, like “send my bottom keyboard to the organ and split the top one at middle C,” I could set up my own rig without asking for sighted help.

    What I’d build first: when I switch setups, the box plays a short sound so I know by ear which one I’m on. I’d also want an “all notes off” footswitch for when something sticks mid-song. And whatever the AI builds should get read back to me in plain text, so I can check the routing before I play.

    On the firmware question, keep the core protected. Whatever I build on top, I want the unit to keep working no matter what. A custom layer on top sounds like the right balance.

    I’d be glad to help test this from a screen reader user’s side.

  • I’m going to add one of your cme widi “zigbee-like” boards to a rpi2040 based project. I hope it only needs coding on the rpi side but if the code inside the board itself could be “tweaked” maybe some nice ideas could arise such implementing OSC… Zigbee itself had some of these ideas implemented because its nature.
    I have a lot of ideas (and some knowledge which makes my live coding not building from scratch) but I think your products could benefeit from community takes like this. I don’t want to solder smd and it seems ok to me paying for the hardware part so maybe some CC by licensing or similar could help keeping projects growing.
    It’s not easy question and I think a lot lately due what’s going on with FM1s, Elektron, etc… and now Behringer/inMusic (Akai) proposals. Maybe it’s not for everyone but surely there’s an untapped market. In fact Korg has some interesting proposals since long ago…

  • It should be opened up in maybe a different product, cause I think this particular product is already highly configurable…also not the novation launchpad mk3 is very configurable to the point of that it can send keyboard commands only if that is what the user wants …. which is great.. just has to be practical .

  • What are the possible downsides, all this could bring is an immense increase of accessibility to extensive midi routing for the general public and is one of the few medians in which AI in music could be extremely beneficial!

  • I want my hardware to work reliability, not for it to be infinitely configurable to my specific wants. It’s not something I am interested in having.

  • In this AI vibe coding era, connectivity and scriptability have become mandatory requirements for my gear. I have no opinion in which way, Sysex, API, or Fw, as long as it supports a generic programming language.

  • I think this will be great. I would absolutely love to be able to configure a custom midi routing / mapping setup for my overgrown midi hardware studio. Back in the late 1980s I learned to write software specifically for that purpose, but don’t want to be tied to a computer all the time. I do think there should be some limits – don’t let just anyone write software to run on hardware if it could damage equipment should it go wrong. So, no electrical power management. But midi data sounds like a safe option.

  • Would be cool if system exclusivity was built in, and acted like a cloud of some kind enabling over midi, the ability to swap data in any direction, like wave files and PCM data. If there was just a device that could store and keep separate protocols for all the different midi compatible devices keyboards, drum machines and so on!. I have the H4MIDI WC built the WIDI in awesome, and the MIDI master kit.

  • Any music related tech which isn’t operated in the creative process should be maximally flexible with minimal work. The less time we spend on routing cables the more we can spend time actually playing. Claude will be the tech that never sleeps, let it set up, maintain, backup and expand on all my HW the way I want it. All HW/FW is already open and out of the makers control, embrace it and provide skills and APIs and ease of integration.

  • Ideas for CME: vibe coding MIDI hardware
    From a musician (jungle and long-form ambient) who builds AI tools that control hardware: MIDI Remote for Cubase, an MCP for the M8 tracker, a SCSI bridge for an Akai S3000XL.

    How open? A layered design
    Locked core: routing, clock, USB host, timing. MIDI’s job is timing; nothing should be able to slow it down.
    Sandboxed user functions on top, each with a fixed time budget and message-rate cap, so a bad function can only fail itself.
    A safe-boot button that always returns to stock. “It broke” is always fixable, which keeps support simple for CME.
    Make it AI-ready
    A published, machine-readable spec for the config and the function API. AI can only write correct code against something it can read.
    A command-line tool or API for HxMIDI Tools. AI can’t work a GUI reliably.
    A simulator to test functions on the computer first. AI code is often “almost right”.
    An approval step: the AI shows a plain-language diff (“port 3: notes above C5 → CC 74”) and the user approves. It never flashes on its own.
    One-tap version history and rollback.
    Explain mode: “here’s what your rig currently does”, in plain words.
    A live MIDI monitor the AI can read, for debugging hung notes and the like.
    Plain-language rules that compile to functions, showing the code so people learn.
    Local-first AI support, with no cloud needed to change your own gear.
    What I’d build
    Clock divide and swing per port, for breakbeat work.
    Tempo-locked humanize, ratchets and Euclidean rhythms.
    A translator layer so vintage synths answer modern controllers.
    Scene switching: one button re-routes the whole rig.
    Probability and slow-drift generators for ever-evolving ambient pieces.
    A SysEx librarian for vintage gear.
    Community and sharing
    A library with readable source, versions and ratings, plus a “CME-verified” badge separate from community uploads.
    Crowd-sourced device profiles (CC maps, SysEx).
    A clear licence for shared functions (e.g. MIT).
    Future-proofing
    A MIDI 2.0-ready function layer.
    An open function format that could work beyond one product, even beyond one brand.
    Accessibility: voice and prompt control for musicians with limited mobility.

    Happy to beta test with a real rig.
    — Thomas Adair

  • I have already done some programming with AI for my Nektar Pacer Midi Pedal and for my Boss GT1000core. I have had Claude program these items using their softwares. It has been great. I currently do not have a need for your products and Ai, however, that could change.

  • The genie is out of the bottle, reverse engineering on the rise and firmware is no longer a black box. I do think open source firmware is the way forward and companies are better invest in open source firmware than futile legal battles around reverse engineering. And the code quality (AI or not) of properly open sourced firmware is very likely better and less buggy than reverse engineered hacking attempts. And I would open source it all, without any restrictions.
    CME could for example open source their firmware and release officially supported and maintained binary packages that can be distinguished (for example via SHA256 or MD5 checksum) from community developed forks.
    Now to the features I would like to program into the H12MIDI Pro. I am using a lot of analog gear that operates with clock pulses instead of MIDI. So my main wish would be a device that combines clock pulses and MIDI. I am assuming that the MIDI inputs and outputs are connected to some I/O pin of the CPU so that my idea is possible, and not an UART or other serial data device. This is what I would probably tell an LLM:
    – add an option that allows to use the TRS MIDI inputs and outputs as clock pulse inputs and outputs.
    – allow different clock pulse formats: 1 PPQN (Roland), 2 PPQN (Korg), 1 Pulse per step. The last option triggers a pulse when a (configurable) MIDI message (CC or Note) arrives.
    – add an individual clock divider to all (analog, MIDI and USB) inputs and outputs. This allows different clock speeds for different devices.
    – add an additional configuration layer “Clock” to the HxMIDI tools software that allows to configure clock routing, dividers etc.

  • While I assume you can add a usbC to Ethernet dongle, an Ethernet port and midi over WiFi would be greatly appreciated. On the software, full MPE support for Zone definition and distribution. KEyboard zone definition and distribution too, should be able to play a room of instruments from one controller.

  • I’m not personally a vibe coder, but being able to modify (or even just having better access to) firmware has the potential to make any relevant thing that much more powerful.

  • I think this is the basis of a good idea but a levels of ‘back to factory settings’ would be essential for those users who make a living from their studios and can’t afford issues when the clients are paying the bill
    My knowledge of AI is limited but obviously it’s a very powerful tool that can prove very useful when used correctly and carefully

  • I’ll give my opinion before my suggestion. For all the downsides of AI, and there are certainly plenty of them, this ability to vibe code software and firmware is fantastic for consumers, and I think for vendors too. Purchase decisions become not just about what the device can do for me, but what it COULD do for me. And I think that’s where the advantage for the vendors is because you can now sell a device to someone who has a concept, rather than purchasing the device for just the features it already has.

    For existing devices, this gives us the ability to have them adapt to our lives, workflows, and preferences.

    This is where the open source advantage comes in. It translates to sales to people who will solve a problem that you cannot or never though to. It also has the advantage of freeing time from your developers to work on the next project.

    In the case of CME, this is even more important because your products focus on solving problems.

    Here is my idea for your software:
    Implement the concept of shift registers for midi mappings, triggered by a note or CC.
    Logic:
    1) Note N on is seen. Enable alternate midi mappings.
    2) Note N off is seen. Disable alternate midi mappings.

    If memory allows, implement an additional alternate.

    Another:
    16 midi maps is very, very insufficient. Look at an Akai Midimix or Novation Launch Control XL. We need 64 mappings per register. (times 2 for one shift register, times 3 for two shift registers).

    Finally, a big thank you for listening and being responsive to customers. This means a lot when added to already great products.

    Cheers,

    Alex.

    • I would also like to have the option to bypass the firmware altogether. Give the ability expose all midi ports via usb. It would make this a tremendous docking station for a daw on a laptop or raspberry pi!

  • I have already made a web audio app, RudimentSeq.com, I’ve just ported that to VCV with a view to hardware version and then now I’m porting it over to the M Wave FM1 platform. So yes it would be very interesting to vibe code for hardware too!

  • Fascinating! I would like to have the feature to diagnose issues and fix them. I would also like to describe all the devices that are connected and then ask to configure it so that the OXI ONE can trigger the drumlogue sequencer whilst playing other tracks on other sequencers, change programs as needed. And for this “prompt” to be savable as a “song”. It should also detail what settings each of the devices need, for example my microfreak is usb connected to my H4Midi and has thru enabled with a midi cable connecting to the Drumlogue.

  • Right now is not about if, but when. The past month thanks to AI:
    – I brought back to life a Tascam US-1800(16 years old) to use on a Silicon Mac (custom made driver, audio and MIDI working again in modern Apple computers)
    – Updated an open source MIDI driver for the MidiSport 4×4 (26 years old) so I can use more than one interface at the same time (the hardware doesn’t have an ID but AI found a workaround, so those boxes are working again in modern Apple computers)
    – Made a fully fledged editor for the AdrenaLinn 3( 19 years old) multi fx pedal (more control than ever now possible)
    And some more projects for “old” hardware are on the queue, so, there!

    If there’s an official path to make things easier to mod or upgrade any piece of hardware, it will be welcome for sure.
    Just add the proper warnings were needed (can something go really wrong on low level code side?)

    About “vibe coding”, actually some knowledge and clear objectives are recommended and needed to speed things up by a lot, but you can always go for the wild adventure if you have lots of time to spare 😀

  • I think that’s a fantastic idea. I want to build presets with macros, transfer functions, gesture recording, sequencing, and whatever else. This would sell the H12 to me. Locking vital processes can be good, but better is explaining dangers and giving tools to understand limits, so those limits may be creatively broken. If the purpose of opening is for vibecoding, then prompting can contain instructions limiting change to anything specific.

  • {"email":"Email address invalid","url":"Website address invalid","required":"Required field missing"}
    >