It’s a well-established fact that a guitarist’s acumen can be accurately gauged by the size of their pedal board- the more stompboxes, the better the player. Why have one box that can do everything when you can have many that do just a few things?
Jokes aside, the idea of replacing an entire pedal collection with a single box is nothing new. Your standard, old-school stompbox is an analog affair, using a combination of filters and amplifiers to achieve a certain sound. Some modern multi-effects processors use software models of older pedals to replicate their sound. These digital pedals have been around since the 90s, but none have been quite like the NeuralPi project. Just released by [GuitarML], the NeuralPi takes about $120 of hardware (including — you guessed it — a Raspberry Pi) and transforms it into the perfect pedal.
The key here, of course, is neural networks. The LSTM at the core of NeuralPi can be trained on any pedal you’ve got laying around to accurately reproduce its sound, and it can even do so with incredibly low latency thanks to Elk Audio OS (which even powers Matt Bellamy’s synth guitar, as used in Muse‘s Simulation Theory World Tour). The result of a trained model is a VST3 plugin, a popular format for describing audio effects.
This isn’t the first time we’ve seen some seriously cool stuff from [GuitarML], and it also hearkens back a bit to some sweet pedal simulation in LTSpice we saw last year. We can’t wait to see this project continue to develop — over time, it would be awesome to see a slick UI, or maybe somebody will design a cool enclosure with some knobs and an honest-to-god pedal for user input!
Thanks to [Mish] for the tip!
62 thoughts on “Neural Networks Emulate Any Guitar Pedal For $120”
Can it emulate the satisfying feeling of unboxing more crap instead of practicing? The ten minutes period where you think this new pedal is the one that will finally make you good before you realize your playing has not improved at all because you were buying pedals instead of playing?
Is there a preset for when you procrastinate so hard you wonder if you even want to play guitar at all, and start experiencing clinical depression symptoms? I think digital is perfectly good for the sound, but can it really capture the experience of hardware?
OK, mea culpa! I laughed so hard it hurt! I need that preset!!!
Good question! It certainly doesn’t DISPLACE all that. In fact, it makes it better. You gotta have the pedal to emulate the pedal. That means you still have to read every article/review, discover THE pedal, hunt it down, order it, feel the anticipation build while you do nothing but track it 3x per day, pace around trying to look you’re do other important stuff, empty the coffee pot several times and finish off the cheesecake until it finally arrives, open it up, see the stage lights and adoring crouds as you walk out and plug it in for the first time, then the whole tail spin when you hit that first chord and the mirage disolves because you still suck and you’re out of coffee and cheesecake and you’ll never get those three days back, so you pull out your computer scroll through all the guitar sub-reddits and repeat. You can still do all that PLUS you get to train the box on this pedal, thinking when the software comes out, you can tweek it and FINALLY get it right.
I mean, not that I’ve ever done anything like that. I’m just imagining…
Maybe if I change pick thickness……nope, still terrible at chords. Oh, the tonewood was the problem, so 3 new guitars later why do all my solos still sound like scales? My amp must be broken because all I hear is string noise and missed notes. This guitar simply will not do pinch harmonic squeals….I need three new guitars. What do you mean it’s excessive to own half a dozen instruments and only know like 3 Ramones songs and the Smoke on the Water riff?????
If you create a single device that does what your saying it will be great for people on a budget and an added plus will be such a large number of musicians will use it that the internet community will be helpful in teaching how to use it
Needs a Blame Shifter, obvs
Can it replicate that tube warmth .the feeling of standing in front of a full stack or the breakup from a good twin .I’m all about building a better mousetrap.if it does the job better
My Fender Mustang V DSP modeling amp does a pretty good job of it with a bunch of DSPs and a 4×12 cab.
Plenty of software VST/AU amp/cab sims used for recording that are absolutely excellent as well.
You are exactly right on this one 👏
Now we have to do it on NE555s…
Hey we got our 555 timer guy!
Where’s the “Handfull of transistors, resistors and diodes” guy. He’s late for the party.
Let’s just use relays instead.
Of course, ironic? Because most overdrive/distortion pedals are just a transistor or diode pair brought into the voltage range where it’s nonlinear…
Still, this is cool.
I’m waiting for someone to do for audio what waifu2x did for images so i can upscale low quality grainy audio to .flac tier quality.
Yeah, that would be extremely cool.
Try Zynaptiq Unfilter and Unchirp… Brace yourself though, the price is a bit hard to digest. Izotope RX also has some useful tools for this, as it’s able to (among other things) reconstruct some high frequency content that was lost from encoding at low sample-rates. Again though, prohibitively expensive if it’s not going to pay for itself in your use case…
I think I’ve matured to the point that I realize not matter how good the pedal is it ain’t going to make my “playing” sound any good
My wife is very appreciative of this type of audio technology because it really does make her sound better when she sings and plays guitar
The standard guitar tuning is meant to work best with male vocalists, so she can change the guitar sound to match her vocal range. She won’t tell me exactly what she’s doing (artists have their secrets) but it certainly works.
Low-tech: Capo; higher-tech: pitch-shifting.
The capo reduces the range of the guitar and requires the user to change keys. DSP can accomplish the same goal without the disadvantages.
NICE but can it emulate the 1980s rockman that gives that Def Leppard sound from the hysteria album or Metallica from all their albums from kill en all till the black album, please keep up the great work, this project will kill tje competition cause theres way to many pedals out there!!!
There’s guitars on the Hysteria album? Sounds like keyboard heavy, chick band stuff to me. Now, if you were to of mentioned, Pyromania…
I do not consider crappy fuzz sounds much of a test. I mean they are the most base of effects. What can it do as fr as emulating something more complicated?
My thoughts exactly. They are grossly overstating it’s capabilities compared to basically any modern digital effects pedal, I’d have to look at the quality of it’s A/D converters, sample rate, and the sophistication of it’s sound engine architecture. I don’t think it can hold a candle
He’s using a HiFiBerry for the A/D and D/A. You _should_ definitely look into those. They hold many candles.
Some kind of overdrive (“crappy fuzz”) is probably exactly what you’d want to showcase a neural-net (read: any nonlinear) modelling system.
The way that distortions/overdrives responds differently to quiet and loud signals is super-important for guitarists. It’s the expression.
All the other other stuff (reverb, delay, filters, cabinet sound, etc) can be handled by linear filters — modelled or impulse-response. It’s the distortions where you need explicit non-linear models.
I agree, it’s all in the “feel”, which makes it hard to evaluate a video like this.
The latency looks ok (hard to say, but it’s low enough that it doesn’t look like it throws his playing off), and it seems to have decent “response” to different levels of soft/hard playing. That is what is really interesting, if it can emulate how a nice tube amp responds to different styles of playing and settings, then it might be a winner.
It would be interesting to hear what a direct recording of the NN output sounds like through good speakers (those computer speakers don’t look confidence inspiring, and it would also cut the room acoustics and camera mic out of the equation).
By the looks of it this is quite impressive. I’d be interested to try it to see how it performs in person.
BTW, it looks like he’s using an HIFIBERRY DAC+ ADC (jumper block on the left seems to match that model). Not sure if the input stage is ideal for a guitar (haven’t looked closely), but at least it can be set to a high sensitivity mode for mic signals. A dedicated mode for instrument signals or a small pre-amp might be good to really match the input signal.
And last but not least, that intro/outro really grates my ears. Tune the guitar FFS…
“Not sure if the input stage is ideal for a guitar (haven’t looked closely), but at least it can be set to a high sensitivity mode for mic signals.”
Like you say, it’s got a high gain mode, and my guess is that the input impedance is pretty darn high. Failing that, toss a JFET in front…
It seems foolish to me to spend a lot of $$ on a pi to get a sound that was pioneered by ripping slits in speakers and hit it’s high point with germanium diodes. You you wanna blow $200 to emulate a ripped speaker or a 27 cent diode, power to you. If it could emulate some of the more fun peddles it might be worth something.
LOL a DS9 Tube Screamer on a solid state amp.
Would be nice to do a Boss Metal Zone MT2 or even something more complicated such as a Morley Evo-1 Delay.
I would love to see a model that makes an electric guitar sound like an acoustic.
If somebody were to put an electric guitar pickup on an acoustic and record both the pickuped and mic’d versions, would this neural network train a decent transform so you could run an electric guitar into the pedal and have an acoustic sounding output?
That would be perfect for me. I have baby skin fingers and don’t play frequently enough to handle the high tension of a nice acoustic, so I play electric. But I long for the sound of an acoustic.
Buy the Yonac iOS app ToneStack Pro (there are other apps). It has an acoustic sim pedal built in.
There are eletronic pedals, that have that effect, like the Zoom G1 four. I can’t name any others, but i’m sure there are others (than Zoom) too.
“Aco.Sim: This effect changes the tone of an electric guitar to make it sound like an acoustic guitar.”
In UbuntuStudio there is a gem packed away with lots of audio video goodies, Rakarack. Hosting 10 of 40 effect “pedals” each with lots of controls and drop downs. So much is adjustable and run more than one instance and branch and do guad or more. You will want experiment rather than just copy one effect style.
Anyone got a Marshall Time Modulator to run this thing on? I’d love to see one of those being emulated. (yes, I know that someone did, once, but they pulled it off the market and AFAIK the thing didn’t really work anyway)
“It’s a well-established fact that a guitarist’s acumen can be accurately gauged by the size of their pedal board- the more stompboxes, the better the player.”
Actually, very often it’s the other way around.
I think that was sarcasm / irony / a kinda self deprecating joke…
Well, I reckon Angus must suck then. I’ve learned over playing the last 40 years the sound is in your hands and the guitar controls, not a pedal. Only one I still use is a wah, only because I can’t reach the tone knob with my pinky while picking. I watched a kids head explode when he cut off the digital delay and I kept playing it 😂. Save your money unless you’re copying some other guy using unecessary pedals, and learn how to do it with your hands. The trick is to play guitar, not effects….
Good point. Im guilty..So many greats did it with a plexi. .or an ac30..I certainly don’t prefer a dry signal ..It may speak volumes about my 37 years of playing however the buffer I get from a nice spring reverb thru a tele is just ear candy to me personally.. I am learning to like tambient effects ..good comment learn to do it with your hands .many greats have said tone is in tge fingers ..but it seems its more found in tge wallet nowadays. ..
This is very cool! I would love to see a more robust comparison of the modeled output against the original training data. An A/B of the same guitar and play through the TS9, then through this NeuralPi pedal, both into some neutral capture device like the HiZ input of an audio interface.
Stop adding gear and pick up your guitar for once.
Is there a neural network that would make me suck less at guitar? The one in my head kind of blows.
Have you tried giving it good training data? Maybe try some kind of pseudo genetic algorithm where successive runs are built upon until the results are satisfactory. Don’t forget to use some kind of positive reinforcement/reward system for progress.
Of course pedals are not scientific constructs, they are works of art. So this device can copy any pedal art that exists thus denying the artist a job,and so degrading any incentive to do art in the first place and providing a means for monopolies to take over the industry, wonderful.
We have already eliminated the possibility of participating in the music industry for nearly all persons, now to eliminate the the pedal industry also, great. FORWARD to human extinction, ra ra technology.
Of course horse drawn carts are not scientific constructs, they are works of art. So this vehicle can copy any cart that exists thus denying the wagoner a job,and so degrading any incentive to ride around town in the first place and providing a means for monopolies to take over the industry, wonderful.
We have already eliminated the possibility of participating in the communications industry for nearly all town criers, now to eliminate the the wagon industry also, great. FORWARD to human extinction, ra ra technology.
Well I live in Russia and I can get an all-tube combo amp for similar price. And I don’t need much of the effects. I’ve had plenty and only kept 2-3. Digital effects are great for recording, sound is great and it’s much cheaper and easier to setup. But they used to have latency and they are not very honest. Like they compression it’s easier to play that way but it’s not very good for practicing. You need to near your mistakes not to hide them.
Community is very conservative too. In 2006 or 2007 I’ve made in-browser tab editor. With export/import to gtp, midi formats. Collaboration like you could share your tabs in public archive. Edit tabs with your mates. Like bassist writes bass party, you guitar and you instantly see changes. I’ve posted a link on some guitar forums and I got reaction “why would we need that, we can pirate guitar pro”. So i gave up. That was before google-docs and similar online services became popular, so maybe it was a bit ahead of it’s time..
Echoplex, Leslie speaker, EMT plate reverb,4ms pedals? will it do guitar synths too?! This could be a fun little box!
I purchased a CXM1978 reverb, and the 3d quality of the tails is not something you can just conjure with a snap of your fingers, let alone the million of different shades that are created by the combination of the reverb’s parameters.
Enjoy hacking away at that kind of challenge!!
But your $899 pedal is a digital emulation model, is it not? Who is to say it can’t be done again?
For starters you also need that kind of physical interface ;-) And then some years to get it to sound “right”. I wish these little boxes – likes the Monome Shield or the PiSound – could help the performing musician or average tinkerer to come up with such effects or sounds on the fly. Even something as powerful as Sueprcollider does require a lot of programming skills to generate sounds, then you need to integrate all the physical controllers (in the case of the CXM, sliders and buttons) via midi or via other proprietary protocols, which have their own quirks, etc.
MIDI controllers are inexpensive these days, the more knobs sliders and other continuous controllers the better!
Am guessing convolution has come a long way since I played with it in HOG and acoustic modeler/mirror in Sound Forge. A friend of mine gave me an Ursa Major SST282 Space Station reverb in 2000 for moving his clothes drier. Was originally billed as the first “affordable” digital reverb.. $1995 in 1978!!
Meant my last reply more to emphasize that in the age of digital, things can advance quickly, but old good stuff can still be good stuff if you use it with skill and taste.
It HAS been done before, after all :-)
When the software will be able to recreate the watery goodness of a nice univibe we’ll talk 😉.
Everybody can get the nuts n bolts… it’s the nuances that make tone be Tone 😉
One pedal alreasy exists that can do all effects. It’s called digital delay. You can get your reverb, flange, auto wah, chorus, etc from a good digital relay pedal.
That’s great how do you get it to change pitch so that half of the human population (female) can sing along in their own range?
Can’t you just tune the guitar up/down? I heard many musicians did that.
All effects you mentioned are only time based effects. You can do all of them by simply playing with time delay on larger or smaller scale. Even comb filter can be done this way. Probably this is why it is called digital delay. Question is if you could emulate let’s say overdriven low pass fillter with resonce or tube amplifier that has gain set to 11?
If someone manages to cast NeuralPi into a MODEP module, the Blokas Pisound with the MODEP emulation will be an instant nice UI and there is also a case for the Pisound, so thats that. https://blokas.io/modep/
Please be kind and respectful to help make the comments section excellent. (Comment Policy)