NAM on Teensy 4.0

guitinker

guitinker

Member
I made a mental note that there was interest in porting the open source Neural Amp Modelling to the Teensy. I can't find that group. It's been a while.

I have a Teensy 4.0 kit that fits inside an electric guitar but I'm not yet using any of the DSP functions at https://www.pjrc.com/teensy/gui/ which I'd like to eventually get to, or find interested parties to be the first to play with it.

I'm currently just sending MIDI PC and CC messages to a laptop running 64-bit VST plug ins, and that works great, but would like to port NAM to the Teensy 4.0 for yet another playground for creatives. That seems to be the direction the wind is blowing. 'Any interest here?

-ARK
 

Attachments

  • 1770409347085.png
    1770409347085.png
    348.1 KB · Views: 67
I have one, working at 55% CPU usage.
Currently still in debugging phase.
I'm rather reluctant to share it since I use LLM and the hatred towards LLM in engineering community is real.
 
Guliediztodies said:
I have one, working at 55% CPU usage.
Currently still in debugging phase.
I'm rather reluctant to share it since I use LLM and the hatred towards LLM in engineering community is real.
Click to expand...
Sadly really it seems that way but I have to be honest I could not achieve what I currently do without it.

For me this is a hobby so happy to use whichever tools to achieve my goals. It’s not exactly an easy path to travel. Sometimes it’s like the blind leading the blind and you end up back tracking.

I do enough challenging coding and engineering to feed and dress my family if I want to team up with Claude or whomever to work on things at night for a bit of lazy code generation then I’m going that way.

Share away if you’re happy to and hopefully this community is kind enough to appreciate the effort you put in
 
bk40 said:
Sadly really it seems that way but I have to be honest I could not achieve what I currently do without it.

For me this is a hobby so happy to use whichever tools to achieve my goals. It’s not exactly an easy path to travel. Sometimes it’s like the blind leading the blind and you end up back tracking.

I do enough challenging coding and engineering to feed and dress my family if I want to team up with Claude or whomever to work on things at night for a bit of lazy code generation then I’m going that way.

Share away if you’re happy to and hopefully this community is kind enough to appreciate the effort you put in
Click to expand...
I will share after getting it right,
Currently I trained a model, the official audio pair for NAM.
I haven't made any script to automatically convert nam into c++ header yet, I directly copied the weights into the project in Arduino IDE.

so far:
1781976456493.png

and , compiling with "fastest + LTO"
1781976980737.png


also, you will need 48Khz sample rate, I use this extension on this repo https://github.com/alex6679/teensy-4-usbAudio , (thanks for the kind person behind it, what a hero you are!)

anyway, what should I do next, what to share? probably training for 44100hz sample rate?(if i can even pull that out...)

Cheers!
*disclaimer: assisted by LLM, use it at your own risk,
 

Attachments

  • NAM_A2_AUDIO_DEBUG.zip
    25.1 KB · Views: 19
You must log in or register to reply here.
Back
Top