The hardware descriptions here seem on the light side. I'd want to be confident that it can handle intense time based effects.
It's promising that they seem to allow arbitrary write to the device, and only charge for tokens for the people that require the prompt playground.
Looking forward to see where this goes.
As an aside: building an ear-pleasing FDN reverb on an obscure-ish board with intense hardware optimization needs has been one of my favorite barometers for the abilities of new LLM models.
I don't know if they did anything fancy like increase the ram or storage or build a custom IC, so YMMV, of course.
> A simple delay could cost $1.00 or $2.00, whereas a complex granular looper might cost up to $5.00.
These prices don't seem reasonable unless there's some really special sauce in their ai.
The real question is: do they have a real-time DSP implementing the AI FX? If not, it's worthless: if you want lag, just use any of the ten million VSTs with your computer; if it is real time, then it's basically a natural language interface on top of an FPGA/DSP -- in which case it could be useful, if it's got reliable presets and pro-grade durability.
I still think most people would take an IR-2, though
But there are other pedals that do custom code, rather than just custom patching, see my other comment https://news.ycombinator.com/item?id=46727231
A pedal you can define with code? Kinda cool, definitely already exists, but kinda cool.
A pedal where you buy tokens to feed the ai monster to generate code to customize your pedal? Ugh. I want off this ride.
Edit - other hackable pedals:
https://www.electrosmash.com/pedalshield
https://www.op-electronics.com/en/dsp-multieffect/696-diydsp...
https://clevelandmusicco.com/hothouse-diy-digital-signal-pro...
Hopefully if the concept catches on more there will be more options for hackable pedals on the market.
Also, those were just the three first ones from memory + a quick google, there are probably other options - but i doubt you’ll find anything under $150. Pedals are not cheap, except when mass produced - and a programmable dsp pedal is by definition niche.
Another option, for people willing to hack a bit, is that the FV1 dsp chip is used in a ton of pedals, and is pretty hackable if you wanted to load your own code: you could find a cheap pedal that uses that chip and hack it.
I don't like these kind of products, what I get in breadth I lose in depth, it's like having a enormous Steam library but only play the first half an hour of each game because I have limited time to invest and too many things. I'm already overwhelmed with my Katana 100.
This is more or less how i learned guitar effects, using cheap digital multi units from ~2005-2010; adding a natural language interface to that doesn't have to be bad, though I'd obviously prefer it explain what it's doing and not just presenting an un-investigable final output. Regardless, there is and always will be a market for beginner guitarists, and at the right price point, i could see this being good for them.
Also, it seems there's no preview in their AI playground, so you have to burn tokens and upload the effect to test it, and it may take lots of iterations to get what you want.
So I think this could mainly interest developers who are able to use it as a platform to develop their own effects without going through the AI thing, and beginners who want to be able to use different community effects to test things.
https://clevelandmusicco.com/hothouse-diy-digital-signal-pro...
Like i said i don't hate the idea. I think it's just a difficult market for this kind of idea.
I play guitar, I've built many pedals, i worked in music retail for the better part of a decade.
I think "one thing does all" pedals are hugely attractive to beginners, with good reason, and I even used to recommend such things to beginners specifically. When i was young, a digital "i can try on every effect ever made!" was an amazing value proposition--assuming the product was cheap enough for my parents to buy it for me. That was usually a Zoom 606 or DigiTech GP50 -- not anything anyone would want to gig with, but an amazing birthday present for a junior guitarist.
Anyway, the reason i never pursued building guitar gear as a job, even though i built plenty of gear i used personally, was mostly because of what i mentioned:
1. Reliability above all else. Charging professional prices for guitar gear means that shit better survive being stomped on 30x/week and a couple a three beers being spilled on in its lifetime.
2. TPB, no question. Use high quality switches, switching caps as needed. Silent, un-loaded transition between on and off is huge -- though, if you're offering delay effects, tails might be desired, the user should have an override, which would require buffered bypass instead. Just be sure to communicate which you use so people know (publishing raw input/output impedance is good; designing to work with vintage impedance-sensitive fuzz, even better)
3. I think most high end guitar players are hyper-picky, and a "Jack of all trades" unit doesn't appeal--the market for that kind of device is decidedly mid-market, and must be priced appropriately to succeed.
It's cool, though. I remember using some GNU audio real-time FX app and plugging my guitar straight into my SoundBlaster and having a great time, despite the unusable lag it all induced.
I think you could capture serious players by offering presets, IR import/copying a given input sound, of course durability, and multi-functionality: is this can replace one effect at a time? Cool I guess, but i have a pedalboard. Can this replace ALL my pedals, including routing and stacking? Vastly more appealing. If it's all software, stacking effects is probably already in the codebas3, and routing would be trivial.
Something like that?
Now, knowing that is something that would sounds cool - that’s the hard part :)
From a cursory glance it appears to be a physical guitar pedal that lets you program virtual effects. The "vibe coding" aspect is likely a system directive + effects library SDK docs fed into an LLM along with the user prompt that generates the appropriate C++ which is then compiled into an effect and run on the pedal.
Note: Which is still very cool. The previous programmable guitar pedals that I've seen were all pretty low-level.
It is still cool though.
I'd LOVE to try it out, or see a demo in person.
My wishlist includes asking the platform to "generate a pedal that sounds like the lead guitar in Comfortably Numb" and it generates that.
I'd pay good $$$ for it.
(It might still not make a Gilmour out of me though :-( )
