KVR Audio

grav3serker · Post by **grav3serker** » Wed Mar 29, 2023 10:21 am

Can someone shoot me an updated Discord invite? The one in the pinned post is expired already. Thanks!

northern_fox · Post by **northern_fox** » Wed Mar 29, 2023 10:42 am

grav3serker wrote: ↑Wed Mar 29, 2023 10:21 am Can someone shoot me an updated Discord invite? The one in the pinned post is expired already. Thanks!

It won't let me post a discord link here, but it's https://pastebin.com/yGCRmKsr

grav3serker · Post by **grav3serker** » Wed Mar 29, 2023 10:52 am

AUTO-ADMIN: Non-MP3, WAV, OGG, SoundCloud, YouTube, Vimeo, Twitter and Facebook links in this post have been protected automatically. Once the member reaches 5 posts the links will function as normal.

northern_fox wrote: ↑Wed Mar 29, 2023 10:42 am
grav3serker wrote: ↑Wed Mar 29, 2023 10:21 am Can someone shoot me an updated Discord invite? The one in the pinned post is expired already. Thanks!
It won't let me post a discord link here, but it's https://pastebin.com/yGCRmKsr (https://pastebin.com/yGCRmKsr)

Thanks a bunch!

Lucadp · Post by **Lucadp** » Fri Mar 31, 2023 2:57 pm

Is it normal that when I load a .nam profile in Presonus Studio one, nam needs 35% of cpu?
My pc specs:
Intel i5 3470
8gb ddr3
Win 10
Nvidia gtx 660 2gb
Ssd crucial 1tb

ladron · Post by **ladron** » Fri Mar 31, 2023 6:04 pm

Lucadp wrote: ↑Fri Mar 31, 2023 2:57 pm Is it normal that when I load a .nam profile in Presonus Studio one, nam needs 35% of cpu?

It depends on the profile. Most people are prioritizing quality over playback speed in their captures, though, so yes, they are generally CPU-intensive.

metalifuxx · Post by **metalifuxx** » Fri Mar 31, 2023 6:31 pm

Lucadp wrote: ↑Fri Mar 31, 2023 2:57 pm Is it normal that when I load a .nam profile in Presonus Studio one, nam needs 35% of cpu?
My pc specs:
Intel i5 3470
8gb ddr3
Win 10
Nvidia gtx 660 2gb
Ssd crucial 1tb

The more epochs a profile is trained with, can give higher quality/accuracy to a certain point, which can be determined with the ESR value/graph generated during and after training. The lower the ESR value, the more accurate it is to the original signal. There reaches a point (i.e. a point of diminishing returns) in training where the ESR wont get any lower with a higher epoch count. Usually the more harmonic content an amp has, i.e a higher gain amp or pedal/amp combination, the more it can benefit from a higher number of epochs used, to be more accurate. *EDIT* See Ladron's post below The more eopochs means higher cpu usage in the captured model. There have been options added into the current trainer software version to give “feather” versions that train with less epochs for lighter/lower cpu usage, and a number of users making profiles are offering captures with both feather light versions (slightly lower accuracy/quality) and also higher epoch count for higher accuracy/quality *EDIT* at the expense of using more cpu.

ladron · Post by **ladron** » Fri Mar 31, 2023 7:17 pm

metalifuxx wrote: ↑Fri Mar 31, 2023 6:31 pm The more eopochs means higher cpu usage in the captured model. There have been options added into the current trainer software version to give “feather” versions that train with less epochs for lighter/lower cpu usage

Increasing the number of epochs makes training take longer, but has no impact on how CPU intensive the model is - that is determined by the network architecture. "feather" models are less complex, requiring less CPU to run, but also generally resulting in higher error rates.

zerocrossing · Post by **zerocrossing** » Fri Mar 31, 2023 8:29 pm

I’m looking for a second story man and a demolition expert to break into Sweetwater after hours. We can capture the entire sales floor before they open.

metalifuxx · Post by **metalifuxx** » Sat Apr 01, 2023 12:56 pm

From Scott Corgan:

Hey everyone!
It's been an awesome couple weeks building and beta testing ToneHunt. We've got about 80 models/packs already up there (and thats not even half of what we plan to have up there)!
For those that haven't heard of the ToneHunt initiative, it started as and continues to be an (unofficial) free, open, secure, and collaborative effort (by many people/mods from this FB group) to make NAM models easy to find and easy to share. It will always be free, always be open, always be secure, and always be collaborative (sharing values with NAM itself!).
We're looking for some more beta testers who love NAM and have models they want to make easily available for everyone. Shoot me a message if you're interested in participating with your models!
For others just looking to find models and share them around, you can sign up to be notified when it officially launches at

https://tonehunt.org

MJACau · Post by **MJACau** » Tue Apr 04, 2023 12:36 am

Is it normal that when I load a .nam profile in Presonus Studio one, nam needs 35% of cpu?

It depends on the profile. Most people are prioritizing quality over playback speed in their captures, though, so yes, they are generally CPU-intensive.

That's dumb, most of the 'Tone' is in the IRs. I'm getting stupidly high CPU usage with it.
Overkill on the epochs just for amp colourisation.

ladron · Post by **ladron** » Tue Apr 04, 2023 6:42 am

MJACau wrote: ↑Tue Apr 04, 2023 12:36 am Overkill on the epochs just for amp colourisation.

As I said a few messages up, the number of epochs has nothing to do with CPU utilization at playback.

northern_fox · Post by **northern_fox** » Tue Apr 04, 2023 5:43 pm

MJACau wrote: ↑Tue Apr 04, 2023 12:36 am
Is it normal that when I load a .nam profile in Presonus Studio one, nam needs 35% of cpu?

It depends on the profile. Most people are prioritizing quality over playback speed in their captures, though, so yes, they are generally CPU-intensive.
That's dumb, most of the 'Tone' is in the IRs. I'm getting stupidly high CPU usage with it.
Overkill on the epochs just for amp colourisation.

Like Mike said, epochs and CPU are unrelated - you can prove this to yourself pretty easily (see attached). What will affect the CPU is the model settings - feather, lite, standard, etc. The IR's provide the EQ curve for the tone, but have nothing to do with gain / response / dynamics / frequency response of the amp.

Fannon · Post by **Fannon** » Tue Apr 04, 2023 9:12 pm

It's probably more related to the size of the neural network. Not sure if this is fixed, but it seems that way. The .nam files are all of the same size. If there were different sizes, I'd expect that NAM models with smaller file size also take less CPU

northern_fox · Post by **northern_fox** » Tue Apr 04, 2023 11:32 pm

Fannon wrote: ↑Tue Apr 04, 2023 9:12 pm It's probably more related to the size of the neural network. Not sure if this is fixed, but it seems that way. The .nam files are all of the same size. If there were different sizes, I'd expect that NAM models with smaller file size also take less CPU

It's not fixed - and you're 100% correct. If you look at feather, lite, standard models you'll see the sizes are varied from 80kb to 400kb, and the CPU usage scales accordingly. You can change the parameters for this if you use the code a bit more deeply than what's offered in the colab or GUI versions, which are designed to make things simpler for most users.

MJACau · Post by **MJACau** » Wed Apr 05, 2023 9:24 pm

The IR's provide the EQ curve for the tone, but have nothing to do with gain / response / dynamics / frequency response of the amp.

Colorisation isn't a monolithic it is "gain / response / dynamics / frequency response of the amp"
Anything that deviates from original signal is distortion aka. colorisation.

Neural Amp Modeler NAM amp capture/profile player FREE *May 12,2023 update 0.7.3