Neural Amp Modeler NAM amp capture/profile player FREE *December 14, 2023 update 0.7.7

VST, AU, AAX, CLAP, etc. Plugin Virtual Effects Discussion
Post Reply New Topic
RELATED
PRODUCTS

Post

jamcat wrote: Tue Mar 21, 2023 7:38 pm TONEX captures can be done with or without cabs, and you specify the signal chain during the Tone Model creation process. The TONEX algorithm is somehow able to separate the amp and cab, and you can replace only the amp or cab component with one from a different Tone Model*, if you wish.

Whether or not there is a dynamic element to the isolated cab component of TONEX Tone Models is unclear, but I don't know why anyone would opt to use a static, linear IR for a cab over a dynamic ML model of one, since speakers are moving things with their own dynamic response curves and saturation thresholds. If anything, what I would do is capture both the amp and cab independently with NAM, and use 2 instances, rather than use ML and IR together.

* TONEX only. Not currently supported in AmpliTube 5.
no, cabs don't have dynamic information to them so IRs are preferrable. I've done plenty of testing on real cab vs IR made from the same session and people can't hear the difference, only if its loud enough to hear the room reflections
the cab is a linear filter
tonex had lots of issues capturing direct profiles, at least in the first months since the release
Funkybot's Evil Twin wrote: Tue Mar 21, 2023 8:09 pm The only 2 attenuators/reactive load boxes with a DI output that will accurately capture the sound of the amp are two units that are very difficult to track down from companies I've never even heard of? And the Suhr, only kind of works? You see why I think amp only captures are not the best idea? :wink: :lol:

Who's got that? How is that easier than: point microphone at speaker about from about 2 inches away, move to center of speaker for brighter, and edge of speaker for darker? :hihi:
well, there could be others that I couldnt get my hands on, but either way, using a real cabinet or a suhr followed by a corrective IR would be just as perfect.
or just using reactive load as is for that matter, its fine, just a little brittle
as for miking the cab - well its more complicated, for starters good speakers are hard to track down, and its already been done by Nolly and ML soundlab and whatnot, redwirez etc. they sound great.
I see your point though, I just believe the results are better when you can craft the sound in a modular fashion rather than hoping that the cab in some profile might work fine

Post

I think its kind of humorous that Bogren Ampknob did the all in one rig thing and then people bitched about how they couldn’t separate the cab from amp model. Then changed it so you could and people were happy. Now this again :lol:

You just cant win :shrug: :dog: :lol: :lol:

Someone always wants to complain about something, even if it’s free. Or there is the fanboi investment buyers remorse thing.

In a mix you wouldn’t even know. All i know is that i’ve never had an easier time mixing and fitting multiple guitar tracks into a busy mix when using NAM.

Post

The other very cool part, is with Tonex you are stuck with one training filing and you can’t edit it or add to it. You might not get the exact response for the specific amp sound and style your seeking and there is no way to change that other than to feature request bomb IK to add more.

With NAM you can take the default training file add in your own playing styles and other tones in the middle of the wav file to get the proper response your seeking.
Last edited by metalifuxx on Tue Mar 21, 2023 8:59 pm, edited 1 time in total.

Post

melvee wrote: Tue Mar 21, 2023 8:26 pm
jamcat wrote: Tue Mar 21, 2023 7:38 pm TONEX captures can be done with or without cabs, and you specify the signal chain during the Tone Model creation process. The TONEX algorithm is somehow able to separate the amp and cab, and you can replace only the amp or cab component with one from a different Tone Model*, if you wish.

Whether or not there is a dynamic element to the isolated cab component of TONEX Tone Models is unclear, but I don't know why anyone would opt to use a static, linear IR for a cab over a dynamic ML model of one, since speakers are moving things with their own dynamic response curves and saturation thresholds. If anything, what I would do is capture both the amp and cab independently with NAM, and use 2 instances, rather than use ML and IR together.

* TONEX only. Not currently supported in AmpliTube 5.
no, cabs don't have dynamic information to them so IRs are preferrable. I've done plenty of testing on real cab vs IR made from the same session and people can't hear the difference, only if its loud enough to hear the room reflections
the cab is a linear filter
tonex had lots of issues capturing direct profiles, at least in the first months since the release
I know from experience that IRs sound filtered and veiled compared to a recording of a real cab. So I think there is more going on with a real speaker than simply acting as a linear filter.
THIS MUSIC HAS BEEN MIXED TO BE PLAYED LOUD SO TURN IT UP

Post

metalifuxx wrote: Tue Mar 21, 2023 8:57 pm The other very cool part, is with Tonex you are stuck with one training filing and you can’t edit it or add to it. You might not get the exact response for the specific amp sound and style your seeking and there is no way to change that other than to feature request bomb IK to add more.

With NAM you can take the default training file add in your own playing styles and other tones in the middle of the wav file to get the proper response your seeking.
That's not how training data works.
THIS MUSIC HAS BEEN MIXED TO BE PLAYED LOUD SO TURN IT UP

Post

But hey, use what you want or justify what you paid for. The user base and group grows by the hundreds each day now. The more this thread is visible, the more users making and sharing profiles. Rock on \m/ :cool:

Post

jamcat wrote: Tue Mar 21, 2023 9:01 pm
metalifuxx wrote: Tue Mar 21, 2023 8:57 pm The other very cool part, is with Tonex you are stuck with one training filing and you can’t edit it or add to it. You might not get the exact response for the specific amp sound and style your seeking and there is no way to change that other than to feature request bomb IK to add more.

With NAM you can take the default training file add in your own playing styles and other tones in the middle of the wav file to get the proper response your seeking.
That's not how training data works.
Some very techy knowledgable people in the group are making very high quality captures would disagree with you on that, and the results will prove that. Unless your trying to say Steve is using the same exact tech under the hood as Tonex or other capture profile hardware/software. Join the group if you haven’t already and share your knowledge/opinions. I’m sure the members would love your input and what you have to say :shrug:

Post

A rising tide lifts all boats. I'm all more advancements in modeling where ever they're coming from.

Anyways, what can you tell us about NAM's parametric modeling?

How does it work?
Is it able to accurately model a full amp in any setting?
Is it practical to do?
What's the downside to it?
Are there many parametric models available, and are many people creating them?
THIS MUSIC HAS BEEN MIXED TO BE PLAYED LOUD SO TURN IT UP

Post

That’s a question for Steve and some of the other techy dudes volunteering to constantly improve and optimize it. Submit feature requests or bugs on the github. It’s complete open source, take a look at the code of the capture/profile software if ya want. And he and anyone else will answer that better than i could in that group.

All i know is that if it sounds good, it is good :cool:

Post

jamcat wrote: Tue Mar 21, 2023 8:58 pm
I know from experience that IRs sound filtered and veiled compared to a recording of a real cab. So I think there is more going on with a real speaker than simply acting as a linear filter.

here's 2 blind comparisons of cab vs IR for you, let me know which is which
https://www.dropbox.com/sh/jb7r2srtune9 ... 6bD4a?dl=0

Post

Funkybot's Evil Twin wrote: Tue Mar 21, 2023 6:39 pm I personally disagree on the 'no cabs is better' thing. Why? In large part for reasons you so clearly articulate.
I think it's one of those things that can go both ways. If someone has great mics, knows mic placement, and has a nice acoustically dead room to record things in, you can end up with a very good capture... however it won't sound like an "amp in a room," because it's a recording of an amp in a room via a mic. I've done it many times over the years and there's no way I'd attempt it because I just plain and simply do not have the proper environment and gear to do it right, and most people don't.
Zerocrossing Media

4th Law of Robotics: When turning evil, display a red indicator light. ~[ ●_● ]~

Post

Another cool thing with the group, the profile creators are there and active. It’s not just a database where one person dumps a profile or two and leaves. They often ask if there are requests for different gain/tone settings if they did not include a variety the first time. And, requested or not, they will often update the pack with more captures they do of other settings. There’s many with some amazing amp collections and they ask people which amps they want modeled first etc. The positive good vibes and collab of the NAM community is something not to often seen in the ampsim threads.

Post

I guess when you see all the first time user posts and you see the one thing common is that their minds are blown on first use and then go on to say they wont need to use another ampsim. Kind of speaks volumes. No one is getting paid to say that.

Steven Wilson of Porcupine Tree said he used a free cheap quality crappy ampsim for a track on the latest Porcupine Tree album. Fans do know he experiments and uses a variety of guitar amps/processors/plugins in the studio, but no one would know that fact if he didnt say.

So at the end of the day if the sound inspires you and it makes ya feel good playing it and sounds great in a mix (and many people tend to say it sounds and feels awesome playing NAM standalone/isolated), then people are going to use it. And yes, free and open source does not hurt either.

Post

melvee wrote: Tue Mar 21, 2023 9:28 pm
jamcat wrote: Tue Mar 21, 2023 8:58 pm
I know from experience that IRs sound filtered and veiled compared to a recording of a real cab. So I think there is more going on with a real speaker than simply acting as a linear filter.

here's 2 blind comparisons of cab vs IR for you, let me know which is which
https://www.dropbox.com/sh/jb7r2srtune9 ... 6bD4a?dl=0
Sure. TEST Y is the real cab. TEST N is the IR.
2101 D and 2101 G are the same recording.
THIS MUSIC HAS BEEN MIXED TO BE PLAYED LOUD SO TURN IT UP

Post

metalifuxx wrote: Tue Mar 21, 2023 7:18 pm
zerocrossing wrote: Tue Mar 21, 2023 2:28 pm
carayo wrote: Tue Mar 21, 2023 10:56 am
AdminNAMcom wrote: Sun Mar 19, 2023 5:44 pm This has been mentioned in the thread before, but NeuralAmpModeler.com is now live.

https://neuralampmodeler.com/

This contains a built in librarian web app called ToneHive for upload and download, ratings of tones and discussion. I am developing a protocol to allow for direct communication between the app and the library which will allow for ToneX like patch switching.
Just discovered NAM, mind blown 🤯 . The fact that is open source is just amazing.
One thing I disliked about ToneX is that you’re paying for user-submitted models. No problem in paying for software or paid models, but if a user submits a model for free, you shouldn’t charge for it.
As a newcomer my biggest issue is that I don’t find it easy to tell which models have speakers/cones included and which don’t. Personally I’d prefer just all-in-one tones that capture the whole amp, including cabinet and speakers. Seems to me like the “standard” intended use is with an additional cab speaker IR.
Maybe for model files it could be useful to have a mandatory field/value that specifies what the model actually includes, specifically cab/cabless. As it is, you usually can’t tell just by the file name.
Yesterday I spent hours looking for .nam files and auditioning them, including finding cabinet IRs that worked well with them. Many were overall duds, like a 65 Deluxe Reverb that had a weird distortion thing going on… as if someone had split the signal and sent half of it to a fuzz box and mixed in 20% of it. If I’d been working, I’d have easily made the $150 to pay for Tonex.

I’ve not tried Tonex on its own, but I did try it in Amplitube 5 and it was fairly clear what captures included cab IRs. What wasn’t clear was how to get to that list, after choosing Tonex as a type. From a UX perspective, it’s pretty horrible. They seem to be optimizing things for you to buy more software, not use the software that you’ve already purchased. Though to be fair, Tonex is free to me at this point, as I didn’t buy it.I do own a bunch of other models for Amplitube. Tonex models should just show up like every other amp choice, but maybe with a little badge on the icon that lets you know what it is. If it is a complete capture, it should automatically disable the cabinet.

But you’re right in pointing out that for a paid plugin, there should be some tagging system that automatically prompts you for an IR for models that don’t have one. At least for their captures. Or maybe just choose a default and put a question mark on the IR slot to let you know that there was no captured cab and it’s just taking a guess.

I think the plugin is really just a way for people who own the hardware to have an extra option. It doesn’t particularly seem like a well thought out piece of software. I’m not sure why anyone would use it if they’re not doing a capture.

NAM is great though, and if you want to capture your own gear or have time to wade though what’s been posted, you can get the same, or maybe better, quality captures. It’s bare bones, though. For how small the UI is, it would be nice if it looked more like a preset browser, where you could more easily see your captures and IRs. The noise gate is a bit crappy, if you’re used to one that’s good. It seems to be very picky about what IR you load into it, so I gave up and just used the free NadIR, which is great.

One thing that you totally miss out on when you use something like NAM or Tonex, when you’re using separate IRs you do not model anything that represents how a tube amp reacts to a load, unless you used a reactive load during the capture. I definitely heard this a lot when sifting though .nam files. It might seem trivial, but a tube amp output stage is directly influenced by how the load of a speaker changes based on what it’s doing. Think of it as an active EQ. That’s where the “warm” phrase started out, as a tube amp will roll off high frequencies at louder volumes, where a transistor amp doesn’t behave differently based on the load, thus sounding “colder.” (Of course, this can be approximated in various ways… but that’s another story.) When I use my AxeFX, which also uses IRs for the cabinet section, you see in the cabinet block that there are controls for how it mimics the tube amp stage. It makes a huge difference if you’re trying to get that tube amp sound, though I’ve also loved pure transistor amps. My old Ampeg stereo chorus amp from the 90s had such a great metal sound, and the cleans were glorious. I miss that thing, but I’m so awash in great amp models that I’m sure it’s mostly just me missing those days.
Have you joined the NAM Facebook group and had a poke around and reading how people are capturing? Your ears or brain must be missing something if you don’t think people aren’t doing the same capture process as what Tonex is doing. Many and most people are using loadboxes after the power amp output between the speaker and a lot of people post pics or vids of the process. Yes, not everyone uses the same loadbox, which will result in different speaker impedances. Some “dishonest” people are probably capturing plugins and not noting that it is a plugin as well, just like in the Tonex database. But for the most part the majority people making the high quality captures are using a quality loadb, and most people will include the ESR deviation calculation and amount of epochs for the training output, as well as the graph. Tonex has a set maximum epoch quality resolution at around 288…most NAM user modeling higher gain harmonic rich amp signals are using between 500-1000 epochs. There are some modeling rack preamps now and synergy modules, alone and with a power amp in the same pack. There have been people capturing power amps now too, which can be used in conjunction with preamp models or seperate ampsim plugins that dont have a power amp section. Lots of pedals now too. There quite wealth of audio tech geek talk and discussions, as well as new users asking about the best loadbox to buy and differences between different loadboxes. Also the amount of people posting quality profiles strictly to the Facebook group is growing so fast every day, its very hard to keep up now. Most people are not using those repository download sites i mentioned yet, you really need to be apart and sift through the group on a daily and sometimes hourly basis. It went from about 200 members a month ago to now over 2,500+, and seems to grow by at least 100 new members a day now with new people wanting to capture and share thier real amps. It’s amazing how fast this group grew.

As far as full rigs, amp and cab, there are some, but they determined the ESR quality is way lower when done in combination with a mic. I don’t know whether IK does it this way or an IR is added and baked into the whole preset, but the general capture process is the same…don’t let Tonex fool ya. At the end of the day, its apples and oranges. Some things NAM does better, and there are a few things that Tonex may do better. You still cant beat free and high quality with NAM though.
I don’t know what people are doing, but today I went though all the stuff I found on the Facebook page and I tossed half of it. A third of it was crap, another third was decent, and the final third was really good. I’m not talking bad like, “I don’t like that kind of amp” bad, because I like all good amps in every style, from Dual Run ectifier to Champs. These just sounded… bad. Not even like bad modeling. Just way off balance, or as I mentioned, the Deluxe Reverb sounded like there was a leaked signal of a fuzz box underneath the main signal.

So… what were the bad ones? I have no idea, but a lot sucked and a lot seemed like you’d be better off with a first generation Amplitube or Guitar Rig model. New stuff from Brainworx or Neural DSP plugins sounded a lot better. I checked the cabs to see if maybe I loaded a bad cabinet IR, but the AxeFX sounded fine though the cabinet. (Free Line6 ones I found)

Anyway, I’m just saying that it’s fine for a free plugin, and I’m grateful to have it. I still ended up with a bunch of really cool captures. I’m lucky that this is my job and I got paid to sift though .nam files, but another person who works a 9-5 who’s not lucky enough to be doing music, might want a quick and easy way to get a bunch of good captures. That’s speculation based on the free Tonex models, but I imagine IK’s pretty good, even if some are not all that great.
Zerocrossing Media

4th Law of Robotics: When turning evil, display a red indicator light. ~[ ●_● ]~

Post Reply

Return to “Effects”