NAM: Neural Amp Modeler

[2dor]

Looks like we're getting a pretty cool enhancement soon for NAM.

Introducing Slimmable NAM: Neural Amp Models with adjustable runtime computational cost

I'm excited to share a paper I've recently completed: "Slimmable NAM: Slimmable NAM: Neural Amp Models with adjustable runtime computational cost"The paper was accepted for the 2025 NeurIPS workshop on Artificial Intelligence for Music: Where Creativity Meets Computation; I'll be presenting it...

www.neuralampmodeler.com

"Slimmable" NAM profiles - you can train normal, very accurate models & use a slider in the NAM plugin to trade off accuracy for resource utilisation.

That's awesome for both users and creators:

- users with less powerful machines can figure out where their sweet spot is
- creators can train 1 single profile at the maximum desired quality & let users decide how they want to use it (full size / accuracy or trimm it down)

A new architecture's coming as well, alongside the STANDARD one we've been used to up until now & judging by the paper Steve's published, it looks like it performs slightly better than STANDARD

 

[Chocol8]

I would have strongly preferred optimization so the full models are less processor intensive, and maybe that's still in the works, but this is a pretty good idea and perhaps a lot better than the file conversion the cheap players have been doing.

 

[Ross Davies]

Jut wanted to say, thanks for the GENOME suggestions guys - all noted down on my side!