IK Multimedia TONEX

synthpenguin said:
and set the Helix’s mids to 6.8 in the case of Fenders that don’t have a mid control)
Click to expand...

This is going to depend on if Helix models the mid pot as linear or audio. The log/audio mid pots on a BF amp need to be around 9 to give 6.8k of resistance to match the amps without a mid control. I see a lot of people trying to run just below 7 to match the 6.8, but that will give you less gain and less mids, unless it is a linear pot, or a model of a linear pot.
 
MirrorProfiles said:
New ToneX update seems pretty good, testing it out now.

Lots of long user requested things have been added such as offline reamping (you can reamp separately to training), queued training (so you can leave it overnight) and adding metadata while it's training. Training file is shorter and slightly less annoying. The time to train is faster on Apple Silicon, although I've only tried advanced modelling so far. It'll stop training once the accuracy hits the limit. You can audition direct tones through an IR (although loading IR's is kind of janky).

Curious to see if the advanced training has closed the gap on NAM at all. Pretty awesome how many of these improvements are directly from things offered by NAM. Its raising the bar for everyone else.
Click to expand...
Oh, sweet!

If the accuracy claims are actually legit, I might actually consider getting a physical ToneX unit.
 
MirrorProfiles said:
Yes. Haven’t compared yet though. Hopefully it means they’ll binned off that high cut at the least, but accuracy improvements across the board would be welcome.
Click to expand...
That's cool. I have a few van valen style NAM captures I've made I want to "transfer" to tonex. The problem has been high end going away. I ended up making compensation IRs to go with the captures, but will try and see now if the update has fixed the high end.
 
Well I did a capture of my Mark V, set to channel 3 on Mark IV mode with the bright switch engaged.

Signal chain:
RME UFXII Output 3 ---> Signal Arts Reamp Box ---> Mesa Boogie Mark V input ---> Rupert Neve Designs DI Box ---> Orange PPC412 with V30's as the LOAD ONLY ---> no cab signal coming back to the RME UFXII Input 9

So the real cab acts as the load. But I have a tap between them to get the amp-only signal.

Here are the comparison files:

Dropbox

www.dropbox.com www.dropbox.com

I used the extended time training mode. I began training at 12:54 and by 12:58, it had finished training. THAT IS A MASSIVE IMPROVEMENT OVER THE OLD VERSION!!! I can't overstate it enough. 4 minutes to train.

I have an NVIDIA 4090 GPU, so I am guessing that's how I was able to achieve that.

1756383752324.png


There is still some aliasing, but I'm fairly sure this is improved from older versions.


Let me know if you think A or B is the real amp.
 
Have I been living under a rock? Did you guys know this update was dropping? I found out this AM via the Analgam marketing email announcing all or most of their amps for V2 somehow? Assuming they had to recapture all the amps they must have had access to this update for a while.

Haven’t touched the update but tentatively have to give some love to @IK Multimedia on this one. They’ve issued a series of updates now heavily aimed at addressing user feedback. This is the right way to do it.
 
Listening on my iPad, I don't have a clue which is which. I will check it out later on real speakers, but I am not confident I will be able to tell.
 
EOengineer said:
Have I been living under a rock? Did you guys know this update was dropping? I found out this AM via the Analgam marketing email announcing all or most of their amps for V2 somehow? Assuming they had to recapture all the amps they must have had access to this update for a while.

Haven’t touched the update but tentatively have to give some love to @IK Multimedia on this one. They’ve issued a series of updates now heavily aimed at addressing user feedback. This is the right way to do it.
Click to expand...
I learned of the update the same way then went to IKM’s website. I saw some blurbs from JSad dated April 2025, but that’s it. I’m not exactly sure what this is at this point.
 
Orvillain said:
Well I did a capture of my Mark V, set to channel 3 on Mark IV mode with the bright switch engaged.

Signal chain:
RME UFXII Output 3 ---> Signal Arts Reamp Box ---> Mesa Boogie Mark V input ---> Rupert Neve Designs DI Box ---> Orange PPC412 with V30's as the LOAD ONLY ---> no cab signal coming back to the RME UFXII Input 9

So the real cab acts as the load. But I have a tap between them to get the amp-only signal.

Here are the comparison files:

Dropbox

www.dropbox.com www.dropbox.com

I used the extended time training mode. I began training at 12:54 and by 12:58, it had finished training. THAT IS A MASSIVE IMPROVEMENT OVER THE OLD VERSION!!! I can't overstate it enough. 4 minutes to train.

I have an NVIDIA 4090 GPU, so I am guessing that's how I was able to achieve that.

View attachment 51050

There is still some aliasing, but I'm fairly sure this is improved from older versions.


Let me know if you think A or B is the real amp.
Click to expand...
That's some real progress. Had a brief listen to the samples - A comes across fuller & with additional high-end content. B sounds close, a bit muffled perhaps & a bit leaner. This may also be down to gain staging I guess but I assume that chain's calibrated for unity gain.

My guess would be A is the real amp & B is ToneX.
 
It’s a solid mark against capture tech that you have to reshoot all your caps every time they update the capture engine. Probably not gonna stop us who like to make captures, but your average user isn’t gonna go for that. They REALLY aren’t going to love having to pay for the V2, V3, V4 version of their favorite paid caps again either.

Not a knock against IK at all, or anyone really. It’s just the nature of this digital tech. Gotta resample the source. That takes time, gear, and ultimately money.
 
Chocol8 said:
IK gets a lot of crap, but they have been listening and the updates Tonex has received have been spot on at addressing the issues. A better training process, batch processing, and better high gain captures sounds like they really hit the mark!

I wonder how Kemper land is feeling today.
Click to expand...
Yeah, absolutely things that users were asking for and a huge step in the right direction.

I’d still love to see:

- better management of IR’s
- cloud training
- embedded input level metadata and automatic calibration
- Ability to re-order the queue of training, and to choose advanced or standard per capture. Would also be cool to see how long each model took to train after the fact.

Orvillain said:
Well I did a capture of my Mark V, set to channel 3 on Mark IV mode with the bright switch engaged.

Signal chain:
RME UFXII Output 3 ---> Signal Arts Reamp Box ---> Mesa Boogie Mark V input ---> Rupert Neve Designs DI Box ---> Orange PPC412 with V30's as the LOAD ONLY ---> no cab signal coming back to the RME UFXII Input 9

So the real cab acts as the load. But I have a tap between them to get the amp-only signal.

Here are the comparison files:

Dropbox

www.dropbox.com www.dropbox.com

I used the extended time training mode. I began training at 12:54 and by 12:58, it had finished training. THAT IS A MASSIVE IMPROVEMENT OVER THE OLD VERSION!!! I can't overstate it enough. 4 minutes to train.

I have an NVIDIA 4090 GPU, so I am guessing that's how I was able to achieve that.

View attachment 51050

There is still some aliasing, but I'm fairly sure this is improved from older versions.


Let me know if you think A or B is the real amp.
Click to expand...
Also going with A as the real amp, feels more alive and juicy, especially in the low end and palm mutes. It's "good for 3 minutes" but I was hoping for closer.

Will do my own tests shortly.
 
2dor said:
That's some real progress. Had a brief listen to the samples - A comes across fuller & with additional high-end content. B sounds close, a bit muffled perhaps & a bit leaner. This may also be down to gain staging I guess but I assume that chain's calibrated for unity gain.

My guess would be A is the real amp & B is ToneX.
Click to expand...
In comparisons against my real amps when performing captures, what you’ve described is exactly what I’ve identified as the Tonex fingerprint. It seems to be just slightly thinner and then the top end is slightly attenuated. This is what I’m hoping gets curved over time. You can still kinda pick it out in the clips.
 
If I’m interpreting correctly that you can now throw an IR on when creating/comparing direct captures, that’s a huge workflow win for me.

Trying to dial/fine tune a direct capture during creation especially feels like the Wild West in a lot of ways.
 
Last edited:
Orvillain said:
Well I did a capture of my Mark V, set to channel 3 on Mark IV mode with the bright switch engaged.

Signal chain:
RME UFXII Output 3 ---> Signal Arts Reamp Box ---> Mesa Boogie Mark V input ---> Rupert Neve Designs DI Box ---> Orange PPC412 with V30's as the LOAD ONLY ---> no cab signal coming back to the RME UFXII Input 9

So the real cab acts as the load. But I have a tap between them to get the amp-only signal.

Here are the comparison files:

Dropbox

www.dropbox.com www.dropbox.com

I used the extended time training mode. I began training at 12:54 and by 12:58, it had finished training. THAT IS A MASSIVE IMPROVEMENT OVER THE OLD VERSION!!! I can't overstate it enough. 4 minutes to train.

I have an NVIDIA 4090 GPU, so I am guessing that's how I was able to achieve that.

View attachment 51050

There is still some aliasing, but I'm fairly sure this is improved from older versions.


Let me know if you think A or B is the real amp.
Click to expand...
I’ve done a ton of ToneX captures and the general characteristics between these are great. A sounds like the tone I want and B sounds like it just needs a small clean boost to it to get to A. I’ll say B is ToneX cause it regularly undergains things.

But man this is pretty promising if I did that capture I’d be stoked on the result… and let’s just say I’ve been very un stoked on ToneX capture accuracy in the past for high gain.

Also 4 minutes is crazy. On my 3900x with a 3090 the advanced training was about 25-35 minutes, totally garbage performance for that kind of machine. Keen to see how it goes now.
 


Some really good sounding tones so far in this playthrough. I feel like I’m hearing a much more lively top end on these caps, more or better harmonic content up top.
 
For anyone that was asking about the new releases, they are live in the app now and are trained on the new V2 software.

It's terrible that the product boxes are now showing up. Hopefully they fix that soon and they state they are indeed V2. Fixed

Also, anyone successfully download the Windows version of 1.10.1? This has been fixed.
 
Last edited:
You must log in or register to reply here.

Similar threads

Orvillain
A mild grumble+warning about IK Multimedia & Amplitube 5 Max
18 19 20
Replies
395
Views
47K
riffy
riffy
Back
Top