Skip to content

tripp528/parted-record

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Trained synth simplest form:

Creating EDM has a huge sound design barrier that gets in the way sometimes.

A great way around this can be to utilize splice. However, if you find a synth bass sound you like, transposing can squash some of the beauty of the sample.

You can also try to recreate the sound using a VST and some effects.

Wouldn't it be cool if we just had a very simple optimizer, where each trainable parameter was a knob on the VST / effects? Even a few hundred trainable parameters (TINY for a machine learning model) would provide the ability to closely replicate most synth sounds found on splice for example.

Training would be simple: you simply 'overfit' your model ('synth preset') on a single sample.

Loss function: feed the model the same fundamental frequency (MIDI note) & compare waveforms.

For starters, this could be just an evolutionary algorithm. Nothing even needs to be differentiable.

Current work worth looking into:

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published