MSPrior & RAVE: mixing unconditional and conditional latent encodings (Anthone dataset)

For this video, I've set up a constellation of an MSPrior model along with encoder and decoder units of a RAVE model. While the MSPrior model generates latent encodings unconditionally, the encoder is being fed a synthesized kick drum, that's being triggered regularly, and forwards its encodings to add to the altered signal of MSPrior.

The dataset both the prior and the RAVE model have been trained on is material I produced and released under the artist name Anthone on my label The Weevil Neighbourhood. https://www.martsman.de/showcase/#ant...


MSPrior is a multi(scale/stream) prior model architecture for realtime temporal learning created by Antoine Caillon.

MSPrior on GitHub: https://github.com/caillonantoine/msp...
RAVE on GitHub: https://github.com/acids-ircam/RAVE
nn~ on GitHub: https://github.com/acids-ircam/nn_tilde

To train RAVE and MSPrior models, you can use notebooks for Colab and Kaggle I've written: https://github.com/devstermarts/noteb...