The positive QC thread

[Orvillain]

Okay! Let's do it! The positive QC thread. Where we only say things we like about it.

- I like that the low-end of the captures is more accurate to the real amp you capture than the Kemper was/is.
- I especially like that the captures are low CPU, so you can put a ton of them into a preset and use scenes to emulate a multi-channel amp.
- I like the walkthrough guide that they give you to tell you how to connect the thing up for capturing. Makes it super easy for people.
- I like the touchscreen! I do! Particularly for simple stuff like naming a preset.
- I like the footswitch+encoders. They work really well.
- I preferred the QC tuner to the Helix tuner. *ducks*

 

[TubeStack]

I like its form factor and it suits my needs.

 

[TheTrueZoltan!]

I like it's form factor and it suits my needs.

Damn, you were quicker than me.

 

[Lysander]

The looper is really, really good. Possibly the best i tried from similar devices.

 

[James Freeman]

I especially like that the captures are low CPU,

That's interesting.
The ToneX as player only device must be using an even lower cost DSP.

It would be very interesting to know how a tiny capture which is basically a set of equations is processing the input to approximate the reference output.

 

[TubeStack]

The looper is really, really good. Possibly the best i tried from similar devices.

This post is Jonathan Nathan Cordy-approved.

 

[Lysander]

It would be very interesting to know how a tiny capture which is basically a set of equations is processing the input to approximate the reference output.

I think QC profiles are fully neural-network based; ToneX's definitely are. And yes, those are very cheap to run, in terms of DSP. The real work goes into creating the NN in the first place.

 

[la szum]

You really need to let it go already.

Not happening? Ok. That's what I thought.

 

[James Freeman]

I think QC profiles are fully neural-network based; ToneX's definitely are. And yes, those are very cheap in terms of DSP. The real work is creating the NN in the first place.

Yeah, the resulting 'capture' file is tiny, just a few equations the DSP runs in realtime.
I know we can approximate any waveform with a Fourier series (long polynomial), since captures are time-invariant each sample of the input has a single unique solution at the output which can be described with just ONE long ass equation.
I believe that's what captures are, a single equation.

PS.
Such a shame wasting valuable discussion topic in a thread soon to be deleted.

 

[Orvillain]

Yeah, the resulting 'capture' file is tiny, just a few equations the DSP runs in realtime.
I know we can approximate any waveform with a Fourier series (long polynomial), since captures are time-invariant each sample of the input has a single unique solution at the output which can be described with just ONE long ass equation.
I believe that's what captures are, a single equation.

PS.
Such a shame wasting valuable discussion topic in a thread soon to be deleted.

I think a capture is basically a set of weight and bias data, but not 100% sure.

 

[Orvillain]

You really need to let it go already.

Not happening? Ok. That's what I thought.

Oi! This is the positive thread!!

 

[Lysander]

I think a capture is basically a set of weight and bias data, but not 100% sure.

^^^ This. The principles of ML applied to audio are super well understood by now; none of it is ground-breaking tech.

 

[MirrorProfiles]

I like:

- the choice of amp models, most parameters fit on one page, all the original amp controls, encoders feel good to adjust.

- moving mic position on the touch screen is great.

- no use to me, but the fact it can process 4 musicians at once is quite a cool feature

 

[yeky83]

I think QC profiles are fully neural-network based; ToneX's definitely are. And yes, those are very cheap to run, in terms of DSP. The real work goes into creating the NN in the first place.

I know QC captures run very cheap in terms of CPU, but I've asked someone online and he said the ToneX plugin runs on the hefty side. I don't think they're that cheap to run... can anyone confirm? Compare CPU loads with an instance of HX Native amps vs. ToneX amps maybe?

Did you mean to say you don't think QC profiles are fully NN based?

 

[mbenigni]

 

[Orvillain]

 

[mbenigni]

OK fine, it's funny.

This time.

 

[dk_ace]

 

[Brandon7s]

I know QC captures run very cheap in terms of CPU, but I've asked someone online and he said the ToneX plugin runs on the hefty side. I don't think they're that cheap to run... can anyone confirm? Compare CPU loads with an instance of HX Native amps vs. ToneX amps maybe?

Did you mean to say you don't think QC profiles are fully NN based?

I've not done any empirical testing, but the ToneX plugin has been very low on CPU resource use for me; similar to S-Gear and definitely less than Helix Native, which is already decently efficient.

 

[Lysander]

I know QC captures run very cheap in terms of CPU, but I've asked someone online and he said the ToneX plugin runs on the hefty side. I don't think they're that cheap to run... can anyone confirm? Compare CPU loads with an instance of HX Native amps vs. ToneX amps maybe?

Haven't tried the plugin, but yes, NNs for audio are quite cheap to run - DSP or not.

Did you mean to say you don't think QC profiles are fully NN based?

Yeah. I'm pretty sure they are though, but Neural has always been (intentionally) vague on how their capturing tech works.