Live data from Hacker News

Show HN: SmartGuitarAmp – Guitar plugin made with deep learning

github.com

91–100 of 163 posts

Re: Show HN: SmartGuitarAmp – Guitar plugin made with deep learning

#91
post #82

Earlier quoted context omitted.

A decent tube amp is not significantly more expensive than a decent modeling amp, particularly considering the used market. They're also just as "bulky" as a solid state amp - the heaviest components are the power transformer and drivers. When you start talking about professional gear, the cost of modeling amps is more than tubes. I'd also contest that this is "figured out" - this area is actively researched.

Price and weight are close. About $2500. AxeFx III is about ~6.8 Kg. Between a half and a third of a Plexi head or an AC-30 Head. For a stadium band. It makes the stage setup much easier. No cabinets, mics and low stage volume. Each member gets a rack with two units (backup) and you are done. Sound will be always the same. Less likely to fail. For a small band, you don't need the bulky rack editions. You get the same…

Yes you have more control, it also sounds worse. They have advantages in live reproduction, yet you still see tubes mic'd up in the studio.

And for a small band even if you have a FoH mix, it's going to sound better out of a 40W combo than through whatever PA your vocalist dragged along. But that's just my opinion.

Digital models do a good job in a vacuum (no pun intended). They do a very poor job of modeling how a guitar and FX chain load and drive the preamp of an real tube amplifier (which is really difficult to capture in either white or black box models). That's also a key to getting lush and evocative ranges of tone from a single configuration.

Just speaking as a player, I don't really fiddle with my tube amps all that much. The tone changes come from dynamics of my hands and controls over my pickup volume controls, and it's accentuated by my drive and boost pedals. That kind of play doesn't translate well to modeling amps which don't have the low headroom, lower impedance inputs of a pure tube stage. I don't need 300 different amp models, I need 1 amp model for 300 tones.

Re: Show HN: SmartGuitarAmp – Guitar plugin made with deep learning

#93
post #44

For people who don’t play guitar, this is really commonplace technology now. Engineers worked out how to model the circuit behavior of any amp a while ago. Powerful valve guitar amps have really interesting effects on the waveform particularly when driven hard, but that valve technology is expensive, unreliable, bulky, and very loud when producing some of the most sought after tones. The most popular guitar amps now…

Is latency an issue?

For purpose-made hardware, basically never.

If you're doing all of the emulation and signal processing on a general-purpose computer, then sure, you often run into latency problems, and part of the set up process for those software tools is latency calibration. The one that I use when at my desktop, Guitarix, is very conservative with its latency, defaulting to like 500 milliseconds, I turn it down to 20-30 or so and it still functions on my hardware and you can't tell the difference: it's the same time delay as playing through a speaker 20 feet away.

Re: Show HN: SmartGuitarAmp – Guitar plugin made with deep learning

#94

Earlier quoted context omitted.

Yeah I didn’t get this statement. I only play tube amps, and most guitarists I know only play tube amps. I actually don’t know anyone that plays a modeling amp. Maybe they’re more popular outside of my sphere, but even with popular YouTube guitars players, they all play tube amps. Mostly Fender.

I never saw the point of getting a real tube amp unless you play exclusively in a studio or rehearsal room. The volume required to crank and get the best of it sounds it makes it too loud for almost any residential area. One of the best features of modeling amps is that you can get the sound of a cranked amp set at a high volume, with low/moderate real volume or even with head-phones. Even in a rehearsal room, real a…

Well the point is the tone. Modeling amps have come a long way, but nothing compares to the real thing.

Of course, most people won’t be able to tell the difference, but many musicians tend to prefer analog.

Re: Show HN: SmartGuitarAmp – Guitar plugin made with deep learning

#95
I was actually thinking that something like this in combination with a “reference” preamp would be the key. This is a similar approach that Townsend microphone’s have taken and they’re supposed to be absolutely fantastic.

In any case, it’s total BS that waves, et. al, can charge hundreds of dollars per plug-in when at the end of the day it’s just software, and I was wondering to myself the other day why there isn’t more stuff in the open source space. I’m grateful to have seen this now, and I’d probably be interested in poking around this kind of stuff more. I have no idea how one actually goes about writing a VST or UA or AU plugin, but this is a great start, thanks!

Re: Show HN: SmartGuitarAmp – Guitar plugin made with deep learning

#96
post #91

Earlier quoted context omitted.

Price and weight are close. About $2500. AxeFx III is about ~6.8 Kg. Between a half and a third of a Plexi head or an AC-30 Head. For a stadium band. It makes the stage setup much easier. No cabinets, mics and low stage volume. Each member gets a rack with two units (backup) and you are done. Sound will be always the same. Less likely to fail. For a small band, you don't need the bulky rack editions. You get the same…

Yes you have more control, it also sounds worse. They have advantages in live reproduction, yet you still see tubes mic'd up in the studio. And for a small band even if you have a FoH mix, it's going to sound better out of a 40W combo than through whatever PA your vocalist dragged along. But that's just my opinion. Digital models do a good job in a vacuum (no pun intended). They do a very poor job of modeling how a g…

> They do a very poor job of modeling how a guitar and FX chain load and drive the preamp of an real tube amplifier

Are there ABX tests for digital vs. physical tubes?

Re: Show HN: SmartGuitarAmp – Guitar plugin made with deep learning

#97

In case any of you want to play with this on your mac and need a usb type-c interface to try this out! :) sorry for the plug but it was too relevant. Great job on the plugin! https://www.effintone.com

It looks pretty neat, awesome form factor. It’d be cool to know more details without having to sign up though.

Re: Show HN: SmartGuitarAmp – Guitar plugin made with deep learning

#98

Earlier quoted context omitted.

I never saw the point of getting a real tube amp unless you play exclusively in a studio or rehearsal room. The volume required to crank and get the best of it sounds it makes it too loud for almost any residential area. One of the best features of modeling amps is that you can get the sound of a cranked amp set at a high volume, with low/moderate real volume or even with head-phones. Even in a rehearsal room, real a…

Well the point is the tone. Modeling amps have come a long way, but nothing compares to the real thing. Of course, most people won’t be able to tell the difference, but many musicians tend to prefer analog.

If most people (ie your audience) can't tell the difference, then why go with it?

Re: Show HN: SmartGuitarAmp – Guitar plugin made with deep learning

#99
post #53

Earlier quoted context omitted.

Is latency an issue?

Latency can be an issue if you are playing through a desktop computer or laptop, but its very manageable (for the most part) if you set up your buffers right. I have never ever noticed any latency issues with any physical hardware, even for equipment from years ago.

A lot of hardware audio interfaces were usb2 which demands cpu time as the system has to poll the usb device rather than act on an interrupt. This caused very low latencies to basically never work without dropouts. I can distinguish a 20-30ms latency and above that it becomes too difficult to ply anything meaningful. Even my semi-pro tascam interface could only sample between 11-14ms. Faster cpus over the last 5 years and ultimately Usb3 solved the polling problem but is just now picking up traction in newer devices due to the usb-c connector.

Re: Show HN: SmartGuitarAmp – Guitar plugin made with deep learning

#100
post #57

Earlier quoted context omitted.

Is latency an issue?

Second Yes. I would not play this at a live gig myself, but it does run fine on my Acer laptop with i5 processor, which is not meant for serious audio processing. If you try it and are getting latency issues, try adjusting the sample rate. Also the stand alone app seems to run much slower than the VST plugin, but most people interested in this I expect would use the plugin anyway.

Get a usb3/usb-c audio interface. Focusrite or presonus are the current favorites. Usb3 does not have the polling problem which cactuses dropouts. Another simple fix is switching to a real-time kernel when using jack (big assumption you are using Linux heh)
Post reply on HN