IK goes AI

[zvenx]

Our heroes won't last that long! They did too many drugs in the 60's!

And then there is Keith Richards, who did so many drugs in the 60s and 70s that he'll somehow outlive us all!

One of the medical histories of all time
rsp

[Funkybot's Evil Twin]

Most of the IK captures are with everything at 12:00, or have I missed something? I would think they'd be neutral, but my experience also reinforces the points made above. Different guitars sound different, and yes, there should be meta data for which type of guitar was used.

The IK Captures seem to capture a variety of settings. Example: I went through a bunch of the AC-30 captures the other night and they all sounded different so I think they were trying to get good tones and capturing the amps at various settings.

[Bombadil]

Yeah, you're right, thx. I get misled by the GUI showing everything at 12:00. Best to read the descriptors.
I've also noted the previously mentioned 'flabby bottom' but it's not too bad, at least for me.

[Funkybot's Evil Twin]

Yeah, you're right, thx. I get misled by the GUI showing everything at 12:00. Best to read the descriptors.
I've also noted the previously mentioned 'flabby bottom' but it's not too bad, at least for me.

Yeah, it can be a little confusing. The amp EQ settings you see on the screen are just like applying an EQ plugin into the signal chain. You can put that EQ before the amp sim in the signal path, or post-amp capture in the signal path, but it's just an EQ. It's not capturing the tone controls on the amp itself. It's just there for extra tone shaping.

[Uncle E]

IK needs to add presets to the parametric EQ. That could be used to solve the different guitars issue.

[DaveClark]

You've heard the capture input file, right?

Yes, in fact I did, but that's irrelevant.

Although it may not be obvious to everyone, the neural net doesn't know anything about playing styles.

For the neural net, there is no demarcation line between styles in the capture input file. It is as if there is one playing style that is some homogenized, weird average of all the playing styles and so forth contained in the clip. The fact that this is a neural net means that we have no idea what this average actually is. All we can hope for is that there is enough material for training, and that the result won't deviate too far off from any one particular style, given that we could feed it just about anything.

I should also point out that I said "may" not "will" cause problems if one deviates from what was used in the training clip, which by the way should have been fully documented.

I'm glad to hear that some people have found that the changes that they have made corresponded to changes with the real corresponding amp. This means that, for those particular captures, things went well. This doesn't mean that there won't be problems with other captures. This is a point-by-point process, not at all like having a physical model.

[jamcat]

Amps don't care about playing styles. They respond to the incoming signal based on dynamics, frequency, and level. So to train the model, you want to feed it the broadest range of material you can, to most accurately replicate the full response range of the amp. Nothing more to it than that.

[Peter - IK Multimedia]

^this, pretty much.

Also, anybody who was having issues with the TONEX iOS app, please update ASAP as we implemented a fix for those who experienced the app crashing upon ToneNET access. Thanks!

[jamcat]

Hey Peter, I have read in a few places that TONEX performance is somewhat lacking when it comes to really high gain tones and can't "chug."

Maybe you need to expand the training data to address that?

[Peter - IK Multimedia]

Hey Peter, I have read in a few places that TONEX performance is somewhat lacking when it comes to really high gain tones and can't "chug."

Maybe you need to expand the training data to address that?

Yeah, I've read that too, mostly early on. Seems there are people out there who have captured Tone Models that may convince those folks that it might not be a TONEX problem after all [edit - not saying TONEX is perfect, meaning that tone preference is subjective so finding the right fit for what one considers proper high gain can be challenging sometimes]. I don't endorse this pack specifically but it was easy to find and seems to be one of many examples of TONEX doing high gain quite well:


Of course TONEX will continue to evolve, though, as nothing is perfect and everything can improve.

[raintalk]

Not too far off!
https://m.media-amazon.com/images/I/71S ... 0_ML2_.jpg

https://www.amazon.co.uk/IK-Multimedia- ... 0BV2Z8YYW/

[jamcat]

Of course TONEX will continue to evolve, though, as nothing is perfect and everything can improve.

What about including a nextgen AI-trained noise gate/de-noise in TONEX?

(Or at least a tuner)

[Peter - IK Multimedia]

Of course TONEX will continue to evolve, though, as nothing is perfect and everything can improve.

What about including a nextgen AI-trained noise gate/de-noise in TONEX?

(Or at least a tuner)

I like the Dime Gate via AmpliTube 5, personally

And... yes.

[LGK_Dude]

Questions that hopefully someone can answer before the intro price ends.
- I understand you have to be online to download profiles from tone.net, but do you have to be online in order to play those profiles live? Does it ever need to periodically "phone home?" Question for both ios app and mac.

- For user profiles, is there any chance of them disappearing? I'm assuming its possible for a user profile to be removed from tone.net? If so, will that also remove it from my system as well, either through some sort of online sync or update? Or will I always have access to profiles I've downloaded, even those that were uploaded by other users?

- Anyone have experience transferring Tonex Profiles and presets from their phone / tablet to their laptop, assuming its even possible?

Tonex sounds really good, and I'm considering buying SE and using it in my band. But if I have to be online either for verification, or if downloaded profiles can be removed from my system if they are removed from tone.net, it would make me wary to use in our music.

[jamcat]

Of course TONEX will continue to evolve, though, as nothing is perfect and everything can improve.

What about including a nextgen AI-trained noise gate/de-noise in TONEX?

(Or at least a tuner)

I like the Dime Gate via AmpliTube 5, personally

Well, I'd like to see the TONEX Noise Gate be replaced with an Ai-powered noise reduction that was trained on guitar noise, and removes it completely without touching the guitar signal.

The AT5 noise gates are a no-go for me now that I've begun phasing out AmpliTube 5. Because I like the decoupling of amp and cab Tone Models in TONEX better, and because IK flat out refuses to update the stomps in AmpliTube for 2023 to be more accurate and dynamic. Everything in AmpliTube needs to be redone with machine learning.

So now I'm using stomp models from Nembrini, Audiority, Fazertone, Brainworx, Kuassa, etc. Much better sound.

The Fazertone pedals are particularly interesting, because they are also all modeled with a neural network inference engine. And they are fully modeled pedals with working knobs, not just snapshots! They have a free Klon Centaur pedal.

I've replicated a JC-120 with stereo chorus by putting the TONEX JC-120 amp with cab disabled, followed by the Audiority Chorus Ensembler, followed by a stereo split with a TONEX JC-120 on each side with the amp disabled. It's a complex chain, but it sounds indistinguishable from a real JC-120 now. It's just too bad it can't all be done in AmpliTube 5 as convincingly.

TONEX is great, but I wonder when (or if) IK is ever going to get serious about cleaning up the mess AmpliTube has become, and get the stomps up to a 2023 standard.