Prism is a multiband distortion audio effect that uses a single neural network to apply separate distortion effects (overdrive, fuzz, distortion) to different frequency bands of your audio signal.
Unlike traditional multiband processors, Prism's neural network learns a single complex transfer function with sophisticated band behaviors, receiving per-band conditioning on effect type, gain, and tone settings.
Prism offers control over 8 frequency bands, and users can select separate effect types, gain and tone settings for each
Available effects are overdrive, fuzz, and distortion.
At the core of Prism is a modeling temporal convolution neural network. The three effects are modeled after analog boutique pedals. The overdrive is modeled after a certain royalty member of the overdrive world. The fuzz models a red pedal which in turn was inspired by the famous muff. The distortion effect models a rebel purple IC distortion pedal.
Prism comes with a GUI inspired by multiband eq pedals. At the moment, the GUI is a simple interface that sends OSC messages to an underlying Python runtime that performs the modeling, but we plan on integrating inference into the gui and compile it as a plugin.
Below are some audio demos
All the rows first show the settings for the 8 frequency bands, which comprise per-band effect type (Fuzz, Overdrive, Distortion), Gain (0-5) and Tone (0-5)
Below are Dry and Wet signals.
Note: both Dry and Wet signals here are processed through virtual amplifier with cabinet simulation (Swanky Amp plugin, clean setting)
In the Dry signal it's just the DI instrument recording through the amp, while in the Wet signal the DI recording is first processed with Prism and then sent through the amp.
This is because non-linear distortion effects interact with the amplifier and are almost never used in isolation.
Full demos including those without the amp are found here.
|
||||||||||||||||||||||||||||||||||||||
|
Dry (DI>Amp): amp_bass.mp3 |
Wet (DI>Prism>Amp): amp_bass_wet |
|||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||
|
Dry (DI>Amp): amp_epiano.mp3 |
Wet (DI>Prism>Amp): amp_epiano_wet |
|||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||
|
Dry (DI>Amp): amp_guitar-chords.mp3 |
Wet (DI>Prism>Amp): amp_guitar-chords_wet |
|||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||
|
Dry (DI>Amp): amp_guitar-funk2.mp3 |
Wet (DI>Prism>Amp): amp_guitar-funk2_wet |
|||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||
|
Dry (DI>Amp): amp_guitar-funk.mp3 |
Wet (DI>Prism>Amp): amp_guitar-funk_wet |
|||||||||||||||||||||||||||||||||||||
Prism is currently in development. Check out the GitHub repositories for the latest updates.
Neural Net Repository
www.github.com/return-nihil/Prism
