Hacker News Reader

keyth72on Show HN: SmartGuitarAmp – Guitar plugin made with deep learning

This is correct.

keyth72on Show HN: SmartGuitarAmp – Guitar plugin made with deep learning

I'm new to Hacker News and I might not understand how it's supposed to be used, can you explain what you mean so I can fix for future posts?

keyth72on Show HN: SmartGuitarAmp – Guitar plugin made with deep learning

Absolutely it's clickbait, I'm trying to get some visibility on my open source project.

To answer the question of why it's $600 specifically, it's based on a model of a Fender Blues Jr. Tweed, msrp $600. The models are evaluated to have a 1.5% error to signal ratio, meaning it's 1.5% different from the actual recorded signal, coming from an SM57 microphone.

I really do appreciate the constructive criticism, and well observed points. I'm new to promoting a project and also this Hacker News site. I'm actually floored it's getting this much attention.

The fact it uses deep learning doesn't make it better than traditionally modeled amps, but it does mean that someone with no domain experience for amp modeling (like me) can come in and make a good sounding guitar plugin.

keyth72on Show HN: SmartGuitarAmp – Guitar plugin made with deep learning

Thank you! Any feedback is appreciated if you get around to trying it out.

If you look at the WaveNetVA repository from Aalto University that my plugin is based on, at least one person affiliated with that project works at Neural DSP. I suspect that the QuadCortex uses something like this (maybe a more advanced neural net) combined with other proprietary techniques and application specific hardware. They claim they can do the rig capture in 3 minutes, whereas the models I trained for the SmartAmp took a few hours on a GPU.