(2018-08-29) Sloan Expressive Temperature

Robin Sloan: Expressive temperature. This is a very niche post documenting a technique I think might be useful to artists—especially musicians—who want to work with material sampled from machine learning systems. (Generative Music)

Often, when you’re using a machine learning system to generate material, sampling temperature is a key expressive control.

It occurred to me, deep into some experiments with audio, that it should be possible to change the temperature not just between samples, but during them

To begin, here’s a sample generated The Normal Way, at constant temperature 1.0.

Here’s another sample—same model, same everything—generated by sweeping the temperature from 0.75 to 1.1 and back:

None of this sampling works real-time, at least not outside of Google. You cannot yet turn a literal knob and listen while the sound changes—seeking, adjusting, conducting—but that “yet” has an expiration date, and I think we’re going to see a very new kind of synthesizer appear very soon now

Edited:    |       |    Search Twitter for discussion