Neural audio latent sequencing with RAVE in Pure Data (Martsman Black Plastics Series custom model)

In this video, I'm presenting another technique to stabilize the output of RAVE neural audio models to an exact loop by generating repeatable sequences of signal information/ pseudo embeddings.
This is done via a compact prototype component for latent sequencing, "arseq", that I've written in Pure Data (Source: https://github.com/devstermarts/PD-co.... It generates arrays of values and cycles through these modulated with simple envelopes on every bang sent from a main metro object. To be most accurate, the metro object needs to be configured with samples as key metric.

The model used in this video has been trained on the "Black Plastics" series of tracks I've released in the past years. https://martsman.bandcamp.com

---

RAVE is "A variational autoencoder for fast and high-quality neural audio synthesis” created by Antoine Caillon and Philippe Esling of Artificial Creative Intelligence and Data Science (ACIDS) at IRCAM, Paris.

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

To train models on Colab or Kaggle, you can use these Jupyter notebooks i've set up: https://github.com/devstermarts/Noteb...