Jason, I just watched your teaching of the NAM first,I had a stroke trying to get back on board,then I will get on my computer. Thanks so much means a ton. Stuart
Thanks! I just did my first model thanks to you. Getting the exact length is super important, also putting 24 bits... had to retry exporting the wav a few times to get it working.
Yeah, its VERY precise. Ive helped a few people who were off by 1 sample or they made a stereo output.wav or it was the wrong bit depth. Glad you got it figured out!!
This is great Jason - thanks, and the consolidated links are very helpful. Depending on which DAW you use to generate the output file, it can be tricky to get it 48kHz 24bit mono wav, the exact same length as the training file. But this is key to being able to generate a model file.
Yep. Some DAWs have a "tail" built into them which always adds extra samples. Its best to drop both files into a basic audio editor like Audacity to check both of them before uploading to Collabs.
No problem! Let me know if you ever get one done! Join the Facebook group and post it! You can also ask questions there if you get hung up on something.
Its definitely getting a ton of attention right now! There are a few geeky folks throwing their hats into the ring, as well. Some really cool developments with the local trainer. More to come, Im sure!
Lol I'm back again. :p Question about epochs and google collab. Is there a rough estimate on the max you can do for a capture before it times out? I'd look on the fb group but I ditched fb like five years ago so am kinda the outsider here haha. It seems like most people do 250 to 500, but I've seen some crazy 1000+ epoch captures, wondering when it starts to become diminishing returns etc. I'll probably just try 100 as a starter and go from there. Thanks for all the great vids on this stuff!
I think people have settled on 300 as the optimal choice for quality and wasting time. Collabs is weird and picky about how long something can run but you just create a new email address or use a different web browser and you basically continue with that one until the other one isnt timed out anymore.
Be careful using the 'neuraldsp' word, because NeuralDSP is a brand name, targeted at the same market. GuitarML is not part of the brand NeuralDSP. Writing that for the people who might get otherwise the wrong association.
Great video, Jason! I have a question about applications for this. When grabbing a capture of the initial .wav file through an amp/amp sim, that cooks in the amp settings, right? I'm thinking about how practical it would be to rent some physical amp heads to run through this and start building a catalog. However, this would be pretty limiting since I'd only be capturing each amp at the specific settings I used during the initial capture, right?
Thats exactly right. This is a "snapshot" of the amp at those settings. I wouldnt go crazy renting amps just yet. The software is being developed. Have fun with it for now and kinda get your feet wet. I used an amp sim as an easy example. I plan on using a combination of plugins and hardware to make an amp tone that is fully produced and polished (minus the IR of course). Thats my next thing to try. You should join the Facebook group! Lots of discussion as well as contributions to the project (people sharing their captures, people helping develop the system, etc). If anything, you can track the progress on things and contribute with your suggestions on what it can do better. You also get a bunch of freebies for amp captures to try out!
This is fairly true of profiling as a whole - it's why you'll often see multiple captures of the same amp just with different settings. Kempers etc all have tone controls, but they're more like applying EQ curves than turning tone controls on real amps.
Correct. there are technical elements to the trainer file that allow for precise quantization and alignment. there are also loud and quiet sounds which help train the model to understand the full volume spectrum. you can replace some of the middle portion of the file with your own DI tracks.
What are the important signal levels or where is a good setting to start when recording the file? Adjust amp gain, amp volume, or signal from v1 file? My first try with mesa mark v amp was .000 but the tone was clean and should have a heavy gain sound from ch2. Thanks for all your videos, great job.
I would set up my hardware loop (audio interface output > Reamp box > amplifier > load box > audio interface input) and then play guitar through that system and listen to the amplifier through your DAW to see how it sounds. Then look at how loud your guitar signal is compared to the v1 file. I would adjust your guitar signal so it is similar in volume to the v1 file. Maybe record a loop of yourself playing some gutiar chords and look at that recorded audio and compare it to the v1 file. They should be relatively the same volume. This way, when you play the V1 file through your amplifier, what you are recording back into your DAW is basically going to sound just as distorted as the gutiar chords loop you were playing through that signal chain a moment ago. That should give you a good Output.wav and a good NAM model.
Thanks for breaking the capture process down! I am going to try this. For clarity, do I need to run my DAW output through a re-amp box to go into my physical amp’s guitar input jack? Also, is there a Mac version of the VST3 file?
He is working on the AU version of the profile loader. He doesnt own a mac so its tough for him to implement. As far as the signal, I think its definitely a good idea to use a reamp box. But really, whatever sound you get out of your amps with a regular line in should be reflected in the model so if you can get a good tone out of your amp with a line level signal, go for it!
I play bass and didn't think much of it until I screwed around it. I'm not 100% sure if I did it correctly, but I'm getting the absolute most CHONKY,and FAT SOUNDING BASS I've ever heard before. Truly don't need anything else. This is a godsend, and a Tonex crusher. Tonex must be absolutely pulling their hair out right now lol. Good stuff man..
About normalizing of wav of DI. I think you need to stay with original audio volume or you will need to boost your guitar signal when play through resulted NAM model.
When we normalize the Output.wav (the recording of the distorted guitar) we are not changing the gain staging of the input signal (the DI track). We are only telling NAM to make the output of the amp model louder. You can experiment with this to see what happens. I also believe that newer versions of this system (there have been several updates since this video) will auto-normalize the output.wav already to help standardize the loudness of each amp model you load.
@@АндрейПолевой-я5и I havent normalized the input.wav in this video. The best practice would be to play your guitar straight into the amplifier to see how that sounds, then adjust the gain of the input.wav until it sounds like it makes the same kind of sound as your gutiar. In this video, I just left my DI signal the gain level I normally use it and I left Input.wav the way that it came to me and it worked out fine. The only normalizing I did was with the processed signal.
This is insane, it sounds incredibly close to the original. Do you think it's because you modeled a VST or it doesn't matter? Also what IR are you using in the video?
I used a VST for simplicity in the video but I have done this same process with other real hardware and it is just as accurate. Its incredibly accurate, so long as you do a good job with your gain-staging and setup. The IR in this video is called 05 Ascendants and I dont really remember where I got it. I think it came from a JST amp sim plugin, where it had a bunch of IRs included.
Actually, just figured it out. Took a look and the original location of this IR was here: STL Tonality IRs\Will Putney\3 Album Guitar Tones IRs So I think you can get these IRs by downloading the demo for the STL Tonality - Will Putney amp sim (unless they changed how it works or something since their newest updates).
@@jasonzdora I see, so I guess I have to pay more attention to the gain staging part. I couldn't find the folder nor the IR though...seems like it's not possible to extract the IRs from the complete plug-in. Thanks anyway!
Try downloading a program called VoidTools - Everything. Its a local search system that works way way better than the Windows search function. Just run that program and search for "05 Ascendants IR.wav" to see if you have the file buried deep within some folder thats like a "secret" part of the installer.
Like others, i get this error: RuntimeError: No response activated the trigger in response to blip 1. Is something wrong with the reamp? I did everything shown in the video with the only difference that the amped sound doesnt completely get aligned with the original file. What can i do please? I even deactivated the warp mode when dropping in the sample and the amped sample after flattening it.
I would ask on the discord or the facebook group. I dont have the technical skills to help you, unfortunately. It just worked for me. Check your exported audio tracks and make sure they are the same length. You dont have to align them but they do have to be the EXACT same length. The extra 1 second of silence at the end of the training file is there on purpose. Its there because the system will self-correct the alignment of the impulses at the beginning of the file. You dont need to do any editing besides making sure both files are the exact same length.
@@matenorthThanks for letting me know! That makes sense. You really dont need the warp functionality for this kind of workflow. It would only screw things up.
There is a "standard" v111 file for download (look in my video description). I also have a new video where I show how to make your own custom v111 file. Just check out my channel. Its the latest video. Its also in my "NAM - Neural Amp Modeler - tutorial videos" playlist.
insane. the possibilities! I have a question, and I know you just did an example ‘in the box’ in your video here. but, what is required for a real amp to be sampled? I have a Trainwreck Express amp that I would love to try this with. I have a Hot Plate that has a line out. I also have a Focusrite Scarlett. I keep seeing ‘remap box’ and that the signal feed into the amp needs to be instrument (not line) level, which makes sense. is that what a reamp box will achieve? is there no way to simulate that with a simple audio interface? thanks for the video. just found out about this, and it’s super exciting to see!
A reamp box is meant to turn a line level signal (~500 to 1500 ohms) to a guitar hi-z level (~50k to 100k ohms) which gives the input of your amplifier the kind of signal it is expecting. This gives you a much more "natural" capture as the v1_1_1.wav training file is being performed through your amplifier in its "intended" impedance and level, so long as you set the level correctly. If you dont have a reamp box, the cheaper way to achieve this is to use a DI box in reverse. Or you could try using a tuner pedal that has a buffered input. Usually those will output a signal with a higher resistance than just a regular line level output from your Scarlett. Not sure what a Hot Plate is but I assume that its line level output would suffice for bringing the signal back into your DAW. The only thing to look out for is to make sure that your amplifier has a "load" on it (the speakers are connected and absorbing all of that energy coming from the amp or your Hot Plate is capable of doing the same thing).
I don't have a very good amp to do a profile and share it in NAM but I want to use other people's profiles and I downloaded the Software but it's "empty". How do I access the NAM library? Thank you!
The best ways are to either go to the Facebook group and look for the Files section. OR go to tonehunt(dot)org which is an unofficial repositorty where you can download models.
Awesome!! You did it with the head, is it possible to do the same but with also the cabinet on in the chain? i mean both to get the full tone, because i like a preset in a plugin but i cant turn off the cabinet/IR, Thanks!
This system does not currently support Amp+Cab profiling. You can give it a try just to see, of course! Some folks have been experimenting and theyve gotten great and poor results. Usually, the high end turns out weird.
Likely you are sending too weak of a signal into the amplifier. I would record some regular gutiar DIs of you strumming some chords and loop those. I would play those through whatever setup you are normally playing the NAM training file through (the output of your audio interface > your reamp box > the input of your amplifier > your load box > the input of your audio interface) and I would listen to how those looped chords sound coming back into your DAW. If the gain is too low, turn up the output on your reamp box. As long as your DI track is the same volume that your guitar is normally when you plug your gutiar straight into your audio interface, the gain level you hear coming through that signal chain should be equivalent to how the model will end up sounding.
Honestly, the one you can afford will be just fine. If you cant afford one, try a Boss tuner pedal or anything else with an active buffer when the pedal is in bypass mode. I have cheap to very expensive reamp boxes and they all work just fine. Of course more expensive can mean higher quality but were talking a 5-10% difference. Dont feel gatekept by costs. Just jump in with what you have and see what happens! You will learn something and improve and then learn more and improve more. Go for it!
I suggest taking a screeen shot of the error and uploading your 2 files to a google drive and then sharing that with either the Discord or the Facebook group and see if someone can help you figure it out.
Hi Jason. I tried using the same method as above, but when I reamp on Ableton (using Plugin and freeze and flatten), the reamped signal goes out of sync just a little much making it impossible for me to train. Any help would be appreaciated. its almost like the reamped signal goes out of tempo by a few milliseconds.
What does the error message say? The frozen/flattened reamp signal can be up to 1 second "out of sync" with the training file. Dont worry about lining them up because the system will do that on its own automatically. You dont have to manually re-align anything. The part you have to worry about is getting the length of both audio clips the EXACT same size. Its likely that is what is causing an error.
@@jasonzdora Thanks for replying so soon. I get an error one the final step of the training " No response activated the trigger in response to blip 1. Is something wrong with the reamp? ". This happens after ive input the metadata in the trainer. My length of the output is exactly the same as the wav file. The only issue i can see from watching your video, is the samples not lining up.
@@feoul92 I would share that error and your experience with the facebook group. Theyre going to have a lot more technical know-how and will likely be able to answer your question. Sorry, Im out of town this week and not by my real computer. Im also not geeky enough with NAM to know what any of the errors mean. Im just a caveman who figured out the stuff I put in the video and thats honestly about as far as it goes for me :)
The process you are talking about (extracting the IR from a guitar amp suite plugin) is a different process. I have a video that shows how to do that. I would not use NAM to copy an IR from the guitar cabinet of a VST. Video here: ua-cam.com/video/a0kouHUz0YQ/v-deo.html
@@patbrady9992 Ive made a few different videos on the subject. The latest one I made is the easiest and quickest and its 100% accurate so I would say just go with that method :)
The best place at the moment would be the Facebook group. Things are just getting started, so there arent a lot of captures yet but it will grow over time. I used "marshall" as an example. I do not know of anyone who has modeled a Marshall amp yet. I have a JCM800 and I might do this very soon, though!
@@AnimeAsLeaders998 Yeah, it would be Steve or any of the other members. Facebook is weird about how it manages group requests. Should be done shortly!
The same you would with any DAW. You just have to learn how to route audio to one of your audio interfaces outputs, send it through a gutiar amp and then get that back into your DAW.
Same as any other amp modeler. In Ableton, it reports back zero samples. It is about as heavy as any of the modern amp sims, CPU-wise. So it can be played with a 64 sample buffer without issue on practically any computer.
@@jasonzdora I just picked it up. Reaper reports 0 samples too, but I assume that this is wrong. I'm A/B'ing the models against my real amps (Recto, 6505). The models definitely have the character, but it's not quite there yet. But DAMN I got scared about my collection becoming worthless right there!
Im actually confused. So you modeled the emissary amp in this vid and used it as your (desired) amp? Also how would you use other modeled amps that you dont have and want to use?
I just used the Emissary as an example because it was easy to do. You can model anything with this software. Not sure I understand the 2nd question. Can you re-phrase?
@@drippy6941 I would just follow the same steps I did in this video. Play the v1_1_1.wav through whatever you are trying to model and record the output.
Having some issues when I go to train the model, I get the error "No response activated the trigger in response to blip 1. Is something wrong with the reamp?". The reamped signal is 20ish samples behind the V1 input, but when I try to manually shift them to align perfectly and reupload the output, I keep getting the same error. This ever happen to you?
You dont need to manually align the audio. The system will let you be a whole second late and it will figure that part out on its own. Im not sure what the error means. I would suggest asking on the Facebook group. Bunch of smarter guys than me on that group who could probably tell you whats up. The most important part is that your audio files are the exact same length. If you get that part right, you should be good.
@@jasonzdora i think i got that error cause i didnt realise i wasnt waiting for the circular waiting time graphic to complete. the file names appear before they are actually ready and i was starting the process but it was causing an error "is something wrong with the reamp"
Not the exact code but Im sure its something similar. This kind of process is used in a bunch of different applications. Machine Learning is one of the stepping stones for all of AI.
@@jasonzdora It will be interesting to see what products develop and how this affects/changes the established leader positions in the amp simulation market. Particularly Line 6, Fractal Audio, Kemper, Universal Audio, Neural DSP, IK Multimedia, Strymon and budget companies like Mooer, NUX and Zoom.
@@BobGuido Definitely an exciting time to be a guitar nerd! The real question on my mind is how will the UA-camrs who shill for those companies react to something like this? Will they acknowledge it for what it is or will they ignore it because no payday?
@@jasonzdora Most popular UA-camrs will ignore it because there's no paid promotion, though I think if the development continues and the features people want in NAM are built quickly enough, the amp simulation manufacturers will have to get more creative with the products they offer. The democratization of the recording industry is proof that if something can be done well enough for free, people are no longer willing to pay for it.
I would check with the Facebook group or the Discord group, as I suggested in the video. They will want you to share your two audio files and what the error message was.
@@jasonzdora I just manage to get it right!! (Apparently) It's training right now as I'm writing this. I sent this message minutes before my third try here in the PC and this time I swithched from Firefox to Google Chrome and it's all doing ok for the moment!
@@jasonzdora I can upload the files fine, but I don't appear to be getting the same results as you were running the first step. At the end of script after running it, I am getting The tensorboard extension is already loaded. To reload it, use: %reload_ext tensorboard
@@satch4u3 OK. Sounds like there is something going on with the Google Colabs website. I dont know how to reload the tensorboard. You might want to post this info in the NAM facebook group and hopefully one of the more techy guys can help out. Sorry I dont have a direct answer for you but I know there are guys there that understand this stuff way better than I (barely) do!
@@jasonzdora I finally got it working. My issue was, when exporting my output wave file, Reaper was set to leave a 1000ms tail on the file. Once I turned that feature off in Reaper, the training started working!
@@satch4u3 Ah! Aweosme :) Yeah, there are some rules for the program that are meant to be restrictive in specific ways to help syphon the possible errors that occur into specific tranches. This is one of them (file lengths need to be EXACTLY the same length). Glad you got it working!
Currently, there are no "presets" but there are a couple projects going on right now where people are building databases on websites for folks to upload/download .nam models. Check out the Facebook group for the current info. This is an open source system so its not as "professional" as the stuff you buy. Since its free and since its built by people in their spare time, its not going to be as easy and accessible as a paid product, unfortunately. With some digging, its not hard to find stuff, though!
Not sure what the "it" is in your statement. Have you created your own NAM model? Or did you try using someone elses NAM model? You might want to ask for help on the Facebook group or the Discord. I would need a lot more details and UA-cam comment section is kind of difficult to have a detailed conversation. The guys on Discord and Facebook are super helpful!
@@Gameup2332 I asked you questions but you didnt answer them. I still dont understand. It sounds like you are using the plugin but you havent loaded any NAM models but I cant be sure.
You have to go to this page and scroll down to the area that says "assets" then you download either the Win.zip or the Mac.dmg depending on your operating system. Then you run the installer. Then you scan for new VST3's in Ableton. Then you look for the folder in the VST3 plugins called "Steve Atkinson" and it should be there. If it isnt, you may need to do a full re-scan on Ableton. For Windows, you hold down ALT and hit the "rescan" button (not sure how to do that on Mac). Let me know if that works!
Hey mate, I dont understand what you are doing here :( I tried the steps in this video: ua-cam.com/video/KHIORngbzJE/v-deo.html And I can use NAM with no problems after that (I can use the captures that other users share), so what is this stuff that you are explaining here? also, what is that running locally thing? what advantages would that have? the plugin is working just fine for me after installing it as a normal software. I do appreciate the effort that you put on these videos but I don´t understand what is the point of this process (this one and the running locally video).
Amner's video is showing you how to use the plugin. My videos are showing you how to use the Training Algorithm in order to create one of the NAM Models that you can download. My videos show you how to make a NAM Model out of your hardware.
@@jasonzdora Ohh that's awesome, I just couldn't understand what this was haha I would like to capture my amps soon so this video should be super helpfull, but after seeing how difficult it is, I'm not sure if I'm doing it hahaha Thanks!
Jason, I just watched your teaching of the NAM first,I had a stroke trying to get back on board,then I will get on my computer. Thanks so much means a ton. Stuart
Its a lot to take in, thats for sure!
Hopefully you recover quickly and give it a shot! :)
Thanks! I just did my first model thanks to you. Getting the exact length is super important, also putting 24 bits... had to retry exporting the wav a few times to get it working.
Yeah, its VERY precise. Ive helped a few people who were off by 1 sample or they made a stereo output.wav or it was the wrong bit depth.
Glad you got it figured out!!
This is great Jason - thanks, and the consolidated links are very helpful. Depending on which DAW you use to generate the output file, it can be tricky to get it 48kHz 24bit mono wav, the exact same length as the training file. But this is key to being able to generate a model file.
Yep. Some DAWs have a "tail" built into them which always adds extra samples. Its best to drop both files into a basic audio editor like Audacity to check both of them before uploading to Collabs.
Thanks for doing this! I never would have figured it out on my own....I really appreciate it.
No problem!
Let me know if you ever get one done! Join the Facebook group and post it! You can also ask questions there if you get hung up on something.
Dude I just deleted my account I didn't know what I was getting into to much for me ..🤣🤣🤣🤣
This is great! Thanks for doing this demo!
You bet! You did a TON of work on this and it deserves to be promoted!
thanks dude!!! I'm gonna have a lot of fun with this
Oh fuck yeah!!! Let me know if you have any questions, brother!
Thanks Jason for the walkthrough. This is great and let's hope the community continue to provide equal support to this project.
Its definitely getting a ton of attention right now! There are a few geeky folks throwing their hats into the ring, as well. Some really cool developments with the local trainer. More to come, Im sure!
Thanks for this. I was easily able to create an amazing rig capture on the first try, zero fuss.
Awesome!! Have fun!
I can't believe how good the "full rig" capture is. Amp into Cab sounds basically perfect. No IR needed, just the profile. Amazing.
Thanks dude. Just tried it and it works great. Fantastic tutorial.
Thanks a lot for this video!
You got it! Hope it helped!
Thanks so much for doing this!
For sure! Join the Facebook group and feel free to check out the latest contributions! Some good amps in there :)
Great stuff man !!!
Very cool Jason!!!
I watched this and I am completely lost . Is everyone getting this ??? Wow …totally lost after the record a guitar and match the length
Thank you!
Lol I'm back again. :p
Question about epochs and google collab. Is there a rough estimate on the max you can do for a capture before it times out? I'd look on the fb group but I ditched fb like five years ago so am kinda the outsider here haha.
It seems like most people do 250 to 500, but I've seen some crazy 1000+ epoch captures, wondering when it starts to become diminishing returns etc. I'll probably just try 100 as a starter and go from there.
Thanks for all the great vids on this stuff!
I think people have settled on 300 as the optimal choice for quality and wasting time. Collabs is weird and picky about how long something can run but you just create a new email address or use a different web browser and you basically continue with that one until the other one isnt timed out anymore.
Great video, but, what if you test it out with actual music rather than noise?
That's the funniest wav file I've ever listened to
The training file? Yeah... its... um.. intereting! :)
Incredible
Its pretty awesome!
GuitarML is also neuraldsp modeller and trainer. Check their Proteus vst
Yep! Gave GuitarML a shoutout in the video. Im looking forward to comparing both of them :)
Be careful using the 'neuraldsp' word, because NeuralDSP is a brand name, targeted at the same market. GuitarML is not part of the brand NeuralDSP.
Writing that for the people who might get otherwise the wrong association.
insane :O
Thanks mate..
Join the Facebook group! Theres a lot more cool new stuff thats been added since this video came out!
Is it possible to capture my Precision Drive pedal with only an interface? I have the SS2+ but no loadbox
Yeah, thats a line-level signal so you wouldnt need a load box. Just go for it!
@@jasonzdora sweet! thanks
Great video, Jason! I have a question about applications for this. When grabbing a capture of the initial .wav file through an amp/amp sim, that cooks in the amp settings, right? I'm thinking about how practical it would be to rent some physical amp heads to run through this and start building a catalog. However, this would be pretty limiting since I'd only be capturing each amp at the specific settings I used during the initial capture, right?
Thats exactly right. This is a "snapshot" of the amp at those settings. I wouldnt go crazy renting amps just yet. The software is being developed. Have fun with it for now and kinda get your feet wet.
I used an amp sim as an easy example. I plan on using a combination of plugins and hardware to make an amp tone that is fully produced and polished (minus the IR of course). Thats my next thing to try.
You should join the Facebook group! Lots of discussion as well as contributions to the project (people sharing their captures, people helping develop the system, etc). If anything, you can track the progress on things and contribute with your suggestions on what it can do better. You also get a bunch of freebies for amp captures to try out!
@@jasonzdora Thanks for the detailed response! I sent a request to the group to join. 🙂
This is fairly true of profiling as a whole - it's why you'll often see multiple captures of the same amp just with different settings. Kempers etc all have tone controls, but they're more like applying EQ curves than turning tone controls on real amps.
Wow!
So, the code requires the prebuilt trainer, correct? You can't train using just any DI track?
Correct. there are technical elements to the trainer file that allow for precise quantization and alignment. there are also loud and quiet sounds which help train the model to understand the full volume spectrum. you can replace some of the middle portion of the file with your own DI tracks.
What are the important signal levels or where is a good setting to start when recording the file? Adjust amp gain, amp volume, or signal from v1 file? My first try with mesa mark v amp was .000 but the tone was clean and should have a heavy gain sound from ch2. Thanks for all your videos, great job.
I would set up my hardware loop (audio interface output > Reamp box > amplifier > load box > audio interface input) and then play guitar through that system and listen to the amplifier through your DAW to see how it sounds. Then look at how loud your guitar signal is compared to the v1 file. I would adjust your guitar signal so it is similar in volume to the v1 file. Maybe record a loop of yourself playing some gutiar chords and look at that recorded audio and compare it to the v1 file. They should be relatively the same volume. This way, when you play the V1 file through your amplifier, what you are recording back into your DAW is basically going to sound just as distorted as the gutiar chords loop you were playing through that signal chain a moment ago. That should give you a good Output.wav and a good NAM model.
is possible to use (for example) Emissary with Nadir with an IR loaded? would like to make a full rig
There is now an ability to make a full rig, yes!
Thanks for breaking the capture process down! I am going to try this. For clarity, do I need to run my DAW output through a re-amp box to go into my physical amp’s guitar input jack? Also, is there a Mac version of the VST3 file?
He is working on the AU version of the profile loader. He doesnt own a mac so its tough for him to implement.
As far as the signal, I think its definitely a good idea to use a reamp box. But really, whatever sound you get out of your amps with a regular line in should be reflected in the model so if you can get a good tone out of your amp with a line level signal, go for it!
I play bass and didn't think much of it until I screwed around it. I'm not 100% sure if I did it correctly, but I'm getting the absolute most CHONKY,and FAT SOUNDING BASS I've ever heard before. Truly don't need anything else. This is a godsend, and a Tonex crusher. Tonex must be absolutely pulling their hair out right now lol. Good stuff man..
Its awesome! And Ive added a bunch of bass preamp models to my Google Drive. I will put a link to it in the video description :)
About normalizing of wav of DI. I think you need to stay with original audio volume or you will need to boost your guitar signal when play through resulted NAM model.
When we normalize the Output.wav (the recording of the distorted guitar) we are not changing the gain staging of the input signal (the DI track). We are only telling NAM to make the output of the amp model louder. You can experiment with this to see what happens. I also believe that newer versions of this system (there have been several updates since this video) will auto-normalize the output.wav already to help standardize the loudness of each amp model you load.
I wrote about input.wav
It's problem for all software ampsims. On some of them help tells recomended input signal volume. As usual -18db
@@АндрейПолевой-я5и I havent normalized the input.wav in this video. The best practice would be to play your guitar straight into the amplifier to see how that sounds, then adjust the gain of the input.wav until it sounds like it makes the same kind of sound as your gutiar. In this video, I just left my DI signal the gain level I normally use it and I left Input.wav the way that it came to me and it worked out fine. The only normalizing I did was with the processed signal.
@@jasonzdora got it. Sorry i thought you make an input signal higher
@@АндрейПолевой-я5и No worries! Im glad you asked and then clarified. Sorry I misunderstood you at first :)
This is insane, it sounds incredibly close to the original. Do you think it's because you modeled a VST or it doesn't matter? Also what IR are you using in the video?
I used a VST for simplicity in the video but I have done this same process with other real hardware and it is just as accurate. Its incredibly accurate, so long as you do a good job with your gain-staging and setup.
The IR in this video is called 05 Ascendants and I dont really remember where I got it. I think it came from a JST amp sim plugin, where it had a bunch of IRs included.
Actually, just figured it out. Took a look and the original location of this IR was here:
STL Tonality IRs\Will Putney\3 Album Guitar Tones IRs
So I think you can get these IRs by downloading the demo for the STL Tonality - Will Putney amp sim (unless they changed how it works or something since their newest updates).
@@jasonzdora I see, so I guess I have to pay more attention to the gain staging part.
I couldn't find the folder nor the IR though...seems like it's not possible to extract the IRs from the complete plug-in. Thanks anyway!
Try downloading a program called VoidTools - Everything. Its a local search system that works way way better than the Windows search function. Just run that program and search for "05 Ascendants IR.wav" to see if you have the file buried deep within some folder thats like a "secret" part of the installer.
@@jasonzdora I've just tried with that as well, didn't find anything. It's like they are not accessible from outside, understandably so.
Like others, i get this error: RuntimeError: No response activated the trigger in response to blip 1. Is something wrong with the reamp? I did everything shown in the video with the only difference that the amped sound doesnt completely get aligned with the original file. What can i do please? I even deactivated the warp mode when dropping in the sample and the amped sample after flattening it.
I would ask on the discord or the facebook group. I dont have the technical skills to help you, unfortunately. It just worked for me.
Check your exported audio tracks and make sure they are the same length. You dont have to align them but they do have to be the EXACT same length. The extra 1 second of silence at the end of the training file is there on purpose. Its there because the system will self-correct the alignment of the impulses at the beginning of the file. You dont need to do any editing besides making sure both files are the exact same length.
I solved it by deactivating auto warp mode in the Ableton settings.
@@matenorthThanks for letting me know! That makes sense. You really dont need the warp functionality for this kind of workflow. It would only screw things up.
Awesome video. how did you create the file V1_1_1.wav ? Thank you :)
There is a "standard" v111 file for download (look in my video description). I also have a new video where I show how to make your own custom v111 file. Just check out my channel. Its the latest video. Its also in my "NAM - Neural Amp Modeler - tutorial videos" playlist.
@@jasonzdora Thank you so much :)
@@jasonzdora I got a file named model.nam when I put it in reaper it sounds clean no distorted guitar like in the video :)
@@mialarssen2627 How are you loading the file? When you say "I put it in reaper" what does that mean?
@@jasonzdora I meant I`ve downloaded with NAM in reaper like any .nam file, and it sound clean and not distorted!
insane. the possibilities!
I have a question, and I know you just did an example ‘in the box’ in your video here. but, what is required for a real amp to be sampled? I have a Trainwreck Express amp that I would love to try this with. I have a Hot Plate that has a line out. I also have a Focusrite Scarlett. I keep seeing ‘remap box’ and that the signal feed into the amp needs to be instrument (not line) level, which makes sense. is that what a reamp box will achieve? is there no way to simulate that with a simple audio interface?
thanks for the video. just found out about this, and it’s super exciting to see!
😂 and just like that, almost like youtube was listening to me… ua-cam.com/video/-xHxmveCUEQ/v-deo.html
A reamp box is meant to turn a line level signal (~500 to 1500 ohms) to a guitar hi-z level (~50k to 100k ohms) which gives the input of your amplifier the kind of signal it is expecting. This gives you a much more "natural" capture as the v1_1_1.wav training file is being performed through your amplifier in its "intended" impedance and level, so long as you set the level correctly. If you dont have a reamp box, the cheaper way to achieve this is to use a DI box in reverse. Or you could try using a tuner pedal that has a buffered input. Usually those will output a signal with a higher resistance than just a regular line level output from your Scarlett.
Not sure what a Hot Plate is but I assume that its line level output would suffice for bringing the signal back into your DAW. The only thing to look out for is to make sure that your amplifier has a "load" on it (the speakers are connected and absorbing all of that energy coming from the amp or your Hot Plate is capable of doing the same thing).
Deactivate auto warp mode in the Ableton settings
I'd like to use an own training file as I'm doing acoustic experiments... How can I find informations about that way?
I have a video for making your own v111 training file. Check it out!
I don't have a very good amp to do a profile and share it in NAM but I want to use other people's profiles and I downloaded the Software but it's "empty". How do I access the NAM library?
Thank you!
The best ways are to either go to the Facebook group and look for the Files section. OR go to tonehunt(dot)org which is an unofficial repositorty where you can download models.
@@jasonzdora Thank you very much, Jason!!
Awesome!! You did it with the head, is it possible to do the same but with also the cabinet on in the chain? i mean both to get the full tone, because i like a preset in a plugin but i cant turn off the cabinet/IR, Thanks!
This system does not currently support Amp+Cab profiling. You can give it a try just to see, of course! Some folks have been experimenting and theyve gotten great and poor results. Usually, the high end turns out weird.
I've made a few captures of my amps and all but one turned out to have much less gain. Where am I going wrong?
Likely you are sending too weak of a signal into the amplifier.
I would record some regular gutiar DIs of you strumming some chords and loop those. I would play those through whatever setup you are normally playing the NAM training file through (the output of your audio interface > your reamp box > the input of your amplifier > your load box > the input of your audio interface) and I would listen to how those looped chords sound coming back into your DAW. If the gain is too low, turn up the output on your reamp box.
As long as your DI track is the same volume that your guitar is normally when you plug your gutiar straight into your audio interface, the gain level you hear coming through that signal chain should be equivalent to how the model will end up sounding.
is there any particular reamp box that's recommended?
Honestly, the one you can afford will be just fine. If you cant afford one, try a Boss tuner pedal or anything else with an active buffer when the pedal is in bypass mode.
I have cheap to very expensive reamp boxes and they all work just fine. Of course more expensive can mean higher quality but were talking a 5-10% difference. Dont feel gatekept by costs. Just jump in with what you have and see what happens! You will learn something and improve and then learn more and improve more. Go for it!
i love the concept of NAM
I suggest taking a screeen shot of the error and uploading your 2 files to a google drive and then sharing that with either the Discord or the Facebook group and see if someone can help you figure it out.
Hi Jason.
I tried using the same method as above, but when I reamp on Ableton (using Plugin and freeze and flatten), the reamped signal goes out of sync just a little much making it impossible for me to train. Any help would be appreaciated.
its almost like the reamped signal goes out of tempo by a few milliseconds.
What does the error message say?
The frozen/flattened reamp signal can be up to 1 second "out of sync" with the training file. Dont worry about lining them up because the system will do that on its own automatically. You dont have to manually re-align anything.
The part you have to worry about is getting the length of both audio clips the EXACT same size. Its likely that is what is causing an error.
@@jasonzdora Thanks for replying so soon.
I get an error one the final step of the training " No response activated the trigger in response to blip 1. Is something wrong with the reamp? ".
This happens after ive input the metadata in the trainer. My length of the output is exactly the same as the wav file. The only issue i can see from watching your video, is the samples not lining up.
@@feoul92 I would share that error and your experience with the facebook group. Theyre going to have a lot more technical know-how and will likely be able to answer your question. Sorry, Im out of town this week and not by my real computer. Im also not geeky enough with NAM to know what any of the errors mean. Im just a caveman who figured out the stuff I put in the video and thats honestly about as far as it goes for me :)
@@jasonzdora Oh!
No worries. appreciate all the help. Thank you
@@feoul92 Deactivate auto warp mode in the Ableton settings
Do I capture the speaker IR of the vst amp the same way but bypass the amp and effects?
The process you are talking about (extracting the IR from a guitar amp suite plugin) is a different process. I have a video that shows how to do that. I would not use NAM to copy an IR from the guitar cabinet of a VST.
Video here:
ua-cam.com/video/a0kouHUz0YQ/v-deo.html
Thanks. Is this a better method than used in your other video using pink noise to make Cab IR's?
@@patbrady9992 Ive made a few different videos on the subject. The latest one I made is the easiest and quickest and its 100% accurate so I would say just go with that method :)
Hi Jason, where can I find some of the models that you mentioned on the video (marshalls etc)? thank you
The best place at the moment would be the Facebook group. Things are just getting started, so there arent a lot of captures yet but it will grow over time.
I used "marshall" as an example. I do not know of anyone who has modeled a Marshall amp yet. I have a JCM800 and I might do this very soon, though!
@@jasonzdora great thanks. I am waiting for Steve (??) to approve my request to join facebook
@@AnimeAsLeaders998 Yeah, it would be Steve or any of the other members. Facebook is weird about how it manages group requests. Should be done shortly!
How can I do this with Cakewalk?
The same you would with any DAW. You just have to learn how to route audio to one of your audio interfaces outputs, send it through a gutiar amp and then get that back into your DAW.
This is awesome. How much is the latency of the plugin?
Same as any other amp modeler. In Ableton, it reports back zero samples. It is about as heavy as any of the modern amp sims, CPU-wise. So it can be played with a 64 sample buffer without issue on practically any computer.
@@jasonzdora I just picked it up. Reaper reports 0 samples too, but I assume that this is wrong. I'm A/B'ing the models against my real amps (Recto, 6505). The models definitely have the character, but it's not quite there yet. But DAMN I got scared about my collection becoming worthless right there!
@@jasonzdora And then again, I am damn impressed about NAM. Maybe I should make my own captures and see.
@@lordgraga Awesome :)
How many epochs did you use on the trainer?
What was your ESR score?
@@jasonzdora I was comparing other peoples' captures to amps that I own, so not 100% apples-to-apples. I need a DI box to train models first.
Im actually confused. So you modeled the emissary amp in this vid and used it as your (desired) amp? Also how would you use other modeled amps that you dont have and want to use?
I just used the Emissary as an example because it was easy to do. You can model anything with this software.
Not sure I understand the 2nd question. Can you re-phrase?
@@jasonzdora say I wanted to use an amp that I don’t have (that someone else modeled) how would I do that?
@@drippy6941 I would just follow the same steps I did in this video. Play the v1_1_1.wav through whatever you are trying to model and record the output.
@@jasonzdora sorry I’m not sure you’re understanding. I don’t have an amp at all, would I still be able to use this?
@@drippy6941 You can use the plugin, yes. You can load other peoples amp models that theyve created into the plugin and play guitar through them.
Having some issues when I go to train the model, I get the error "No response activated the trigger in response to blip 1. Is something wrong with the reamp?". The reamped signal is 20ish samples behind the V1 input, but when I try to manually shift them to align perfectly and reupload the output, I keep getting the same error. This ever happen to you?
You dont need to manually align the audio. The system will let you be a whole second late and it will figure that part out on its own.
Im not sure what the error means. I would suggest asking on the Facebook group. Bunch of smarter guys than me on that group who could probably tell you whats up. The most important part is that your audio files are the exact same length. If you get that part right, you should be good.
@@jasonzdora i think i got that error cause i didnt realise i wasnt waiting for the circular waiting time graphic to complete. the file names appear before they are actually ready and i was starting the process but it was causing an error "is something wrong with the reamp"
Deactivate auto warp mode in the Ableton settings
Do I need a reamp box for this?
In general, yes.
Did IK Multimedia just borrow this code for their ToneX plugin and pedal?
Not the exact code but Im sure its something similar. This kind of process is used in a bunch of different applications. Machine Learning is one of the stepping stones for all of AI.
@@jasonzdora It will be interesting to see what products develop and how this affects/changes the established leader positions in the amp simulation market. Particularly Line 6, Fractal Audio, Kemper, Universal Audio, Neural DSP, IK Multimedia, Strymon and budget companies like Mooer, NUX and Zoom.
@@BobGuido Definitely an exciting time to be a guitar nerd! The real question on my mind is how will the UA-camrs who shill for those companies react to something like this? Will they acknowledge it for what it is or will they ignore it because no payday?
@@jasonzdora Most popular UA-camrs will ignore it because there's no paid promotion, though I think if the development continues and the features people want in NAM are built quickly enough, the amp simulation manufacturers will have to get more creative with the products they offer. The democratization of the recording industry is proof that if something can be done well enough for free, people are no longer willing to pay for it.
@@BobGuido Exactly!
the drive file is gone... :(
look up easy_colab.ipynb should be the same
Man, I did the step by step twice already and it got an error message both times.
I would check with the Facebook group or the Discord group, as I suggested in the video. They will want you to share your two audio files and what the error message was.
@@jasonzdora I just manage to get it right!! (Apparently) It's training right now as I'm writing this. I sent this message minutes before my third try here in the PC and this time I swithched from Firefox to Google Chrome and it's all doing ok for the moment!
@@jasonzdora I finally got it!!! Thank you Jason!!
I have tried your process but it doesn't seem to be working. Has the process changed at all?
It hasn't changed. What is the error message? Or where are you getting stuck?
@@jasonzdora I can upload the files fine, but I don't appear to be getting the same results as you were running the first step. At the end of script after running it, I am getting The tensorboard extension is already loaded. To reload it, use:
%reload_ext tensorboard
@@satch4u3 OK. Sounds like there is something going on with the Google Colabs website. I dont know how to reload the tensorboard. You might want to post this info in the NAM facebook group and hopefully one of the more techy guys can help out. Sorry I dont have a direct answer for you but I know there are guys there that understand this stuff way better than I (barely) do!
@@jasonzdora I finally got it working. My issue was, when exporting my output wave file, Reaper was set to leave a 1000ms tail on the file. Once I turned that feature off in Reaper, the training started working!
@@satch4u3 Ah! Aweosme :)
Yeah, there are some rules for the program that are meant to be restrictive in specific ways to help syphon the possible errors that occur into specific tranches. This is one of them (file lengths need to be EXACTLY the same length).
Glad you got it working!
why freeze and flatten?
Thats just the way I did it. You can export if you want. I just flattened so I could show the waveform on screen in the video.
@@jasonzdora thank you for replying. And thank you for your very helpful video!
JSON is pronounced "jay-sawn" lol. It's an incredibly common language/filetype.
Close enough :)
where are the presets ?????
Currently, there are no "presets" but there are a couple projects going on right now where people are building databases on websites for folks to upload/download .nam models. Check out the Facebook group for the current info.
This is an open source system so its not as "professional" as the stuff you buy. Since its free and since its built by people in their spare time, its not going to be as easy and accessible as a paid product, unfortunately. With some digging, its not hard to find stuff, though!
I guess its pronounced "epic" but I kept saying "e-pock". Must be the Chicago accent?
it wont work it just is a clean sound even though i have a impulse and amp thing it wont change
Not sure what the "it" is in your statement. Have you created your own NAM model? Or did you try using someone elses NAM model?
You might want to ask for help on the Facebook group or the Discord. I would need a lot more details and UA-cam comment section is kind of difficult to have a detailed conversation. The guys on Discord and Facebook are super helpful!
wwhat is the discord@@jasonzdora
like my guitar does not change tone even though im changing the settings @@jasonzdora
@@Gameup2332 I asked you questions but you didnt answer them. I still dont understand. It sounds like you are using the plugin but you havent loaded any NAM models but I cant be sure.
do you have discord
@@jasonzdora
Error-signal ratio = 0.893
...Something seems to have gone wrong.
Alright, and what? :D
Noisy signal, usually.
Does anyone have any idea how to install NAM for use in ABLETON? I have tried everything I can figure out. Its retarded.
You have to go to this page and scroll down to the area that says "assets" then you download either the Win.zip or the Mac.dmg depending on your operating system. Then you run the installer. Then you scan for new VST3's in Ableton. Then you look for the folder in the VST3 plugins called "Steve Atkinson" and it should be there. If it isnt, you may need to do a full re-scan on Ableton. For Windows, you hold down ALT and hit the "rescan" button (not sure how to do that on Mac).
Let me know if that works!
Here is the plugin download page:
github.com/sdatkinson/NeuralAmpModelerPlugin/releases
Is this your riff?
No. This is a DI called "lasse and magoo" that I found on the internet.
Hey mate, I dont understand what you are doing here :(
I tried the steps in this video: ua-cam.com/video/KHIORngbzJE/v-deo.html
And I can use NAM with no problems after that (I can use the captures that other users share), so what is this stuff that you are explaining here? also, what is that running locally thing? what advantages would that have? the plugin is working just fine for me after installing it as a normal software.
I do appreciate the effort that you put on these videos but I don´t understand what is the point of this process (this one and the running locally video).
Amner's video is showing you how to use the plugin. My videos are showing you how to use the Training Algorithm in order to create one of the NAM Models that you can download. My videos show you how to make a NAM Model out of your hardware.
@@jasonzdora Ohh that's awesome, I just couldn't understand what this was haha I would like to capture my amps soon so this video should be super helpfull, but after seeing how difficult it is, I'm not sure if I'm doing it hahaha Thanks!
@@ceniam12 Good luck! Be sure to share your models with the community :)
@@jasonzdora sure I will!