Firmware Freedom

Firmware freedom When we released CoyoPedal, we ended with a paragraph about owning hardware and choosing what runs on it, and promised it deserved a post of its own. The discussion that followed on Reddit made it clear that the paragraph was the part people wanted to talk about. So here is the longer version, and an invitation at the end of it. Unlimited tweakability, two sounds I have played guitar for twenty-five years and performed for twenty. In that time I have owned, borrowed and gigged through most of the multi-effects units that mattered, and every one of them sold me the same promise: hundreds of amps, thousands of presets, endless routing, deep menus, a companion app. Here is what actually happens. However long I spend with any of these boxes, I end up on a couple of sounds. One clean, one driven, perhaps a third. That is exactly how it worked with an analogue pedalboard, and it is how it still works with a modeller. I play a three-hour gig on one or two presets and toggle effects on and off as the song asks for them. That is the whole job. The tweakability is real. It is also mostly beside the point for a working player. What decides whether a unit stays on my board is much more mundane: does it have the inputs I need, does it have enough footswitches, does the mic input work because I sing too, does it fit the case, and do the two sounds I use every night actually sound right. The industry has these two questions the wrong way round. It competes on the menu and treats the hardware as the delivery mechanism for it. For most of us it is the reverse. Buy hardware for what it is Take the Ampero II Stage. I like it a great deal. The form factor is right, the footswitches are where my feet expect them, it has the knobs and controls I want, and it has a microphone input, which for a singing guitarist is not a nice-to-have. As a piece of hardware it is close to what I would design myself. It also has a three-core DSP platform inside it with, by our estimate, something like five times the processing headroom of an ESP32. And it cannot run a full-size NAM A2 profile. Not because the silicon cannot do it, but because the firmware was never written to. That is the gap firmware freedom is about. I am happy to pay whatever a good enclosure, good converters, good switches and a good screen cost. What I am not happy about is being told that the sounds, the interface and the workflow of that hardware are fixed by whoever wrote the firmware, forever, whatever my own needs turn out to be. We should buy hardware for what it is: hardware. Then we should be able to run what we want on it, the way we can on a computer, a phone with an unlockable bootloader, or an e-reader with a community firmware. The device neutrality argument is not new. It has simply never been applied to the box under a guitarist's feet. We have already shown what this looks like in a small way. An iRig HD X is a guitar interface with a good input, a headphone output and a volume wheel. Its firmware made it an interface. A different firmware made it an amp modeller, and nothing about the hardware changed. The chips are already good enough, and cheap The other half of this argument is about what a processor costs in 2026. An ESP32-S3 is about $3 as a chip. We have A2-Full running on it live, with six effects and a tuner around the amp, hosting a USB audio interface and drawing a touchscreen. That is the S3 at its limit, and it is the chip CoyoPedal ships on. An ESP32-P4 is about $4. When we first wrote about it it ran one A2-Full model. It now runs two A2-Full models at once, in a fully stereo dual-lane configuration, with ten more effect blocks alongside them. That is a two-amp wet/dry/wet rig on a chip that costs less than a set of strings. Set that beside the dedicated DSPs that the established platforms are built around, and the comparison is not flattering to the DSPs. General-purpose microcontrollers with vector units have caught up on the arithmetic, they are dramatically cheaper, and they come with a modern toolchain, USB host, Wi-Fi and Bluetooth for free. Designing a custom board around one of these that competes with the big platforms on sound is, believe it or not, a no-brainer. We are doing exactly that. There is no secret sauce This is the uncomfortable part for the industry, so let me say it plainly. There is no secret sauce in any of this. The amp models are open: Neural Amp Modeler is open-source, and the capture libraries are enormous and free. The effects are textbook DSP. The hard parts, making it fit a small chip and making it feel like a pedal, are engineering, not magic, and we have been writing up how we did it all summer. Writing a new effect has become genuinely quick, too. If you have taste and know what you are listening for, designing the effect of your dreams is an afternoon or two of work, with modern AI tools doing a good share of the typing. I recently built a physically modelled spring reverb tuned to my own ear that way. Full disclosure: I trained as an electronics engineer and started working on audio circuits twenty years ago, so I had a head start. But it is doable, and we will release that reverb as a plugin soon so you can judge it yourself. When the models are open, the effects are known, and the chips are cheap, what is left to keep proprietary is the firmware. Companies keep it closed because it is the only thing that is still closed, not because there is anything in it that could not be rebuilt. Full circle Fifteen years ago I built pedalboard.js, a library for building guitar effects in the browser with the Web Audio API. It was a toy in the best sense: you dragged pedals around a page and played through them. Today CoyoPedal runs in the browser in much the same fashion, except that the same code also compiles to a $3 microcontroller and runs a real amp model. We have come full circle, and I think the timing is finally right for the thing pedalboard.js was too early to be: an open ecosystem for guitar effects that does not belong to any one manufacturer. We are also building the hardware side of that in earnest. We have partnered with one of the world's most innovative instrument makers to bring a unit built on these principles to life. More on that when it is ready to show. What we want to build, together CoyoPedal is the seed. The goal is for it to grow into a community-driven firmware for guitar processing that can be loaded onto any processor platform and any hardware of the player's choosing. One codebase, with the amp engine, effects, presets and interface, that targets the S3 and the P4 today and whatever DSP or microcontroller you point it at tomorrow. Daisy-chaining boxes is one answer to "I need more"; a firmware that runs the same rig on a bigger chip is a better one. That is not something we can do alone, and we do not want to. So, plainly: - Open-source community. The source is on GitHub under GPLv3. Ports to new boards, new effects, better editors, test coverage, documentation: all of it is welcome, and we will help you get started. - Guitar community. Tell us what your rig actually needs. Not the feature list, the two sounds and the six switches. Build a pedal, capture an amp, write a reverb you like better than ours, and show us. - Processor companies. If you make a chip with a vector unit and a USB host, we would like to run on it. We have done S3 and P4; we are happy to bring the engine up on your silicon and publish the numbers. - Existing pedal and modeller companies. This is the invitation we care most about. You already make the hardware people love. Open it up. Document the codec, the switches, the screen and the DSP, unlock the bootloader, and let the firmware become something the community can improve. You will sell more hardware, not less, and your customers will stop waiting for a model that never ships. Nobody is asking anyone to give away their own firmware. We are asking for the door to be unlocked so that a different one can go in. If any of that is you, get in touch. Firmware freedom is the direction the whole field is heading, and we intend to lead the charge. Welcome to the new world.