Pre-Trained Terrains (Presets)

Make sure to have both:

• an autoencoder (a .ts file)
• a terrain model (a .pt file).

Autoencoder: Percussion - RAVE v1

Pre-trained autoencoder author: Antoine Caillon
Download from RAVE official repository percussion.ts.

• Terrain: percussion-terrain-demo.pt
  Config: 2 inputs 4 outputs, range: -16, -4, 16, 4
  

Autoencoder: Music2Latent-Scripted

Original model is the work by Marco Pasini at SonyCSLParis.
Scripted model can be downloaded from music2latent.ts.

• Terrain: m2l-demo.pt (the one used in the demo video)
  Training data from: Stem from an EDM track I made under the name Alaska Winter.
  Config: 2 inputs 64 outputs, range: -16, -4, 16, 4
  

Autoencoder: Stable Audio Open 1.0

Pre-trained model by Stability AI.
A streamable scripted version for realtime continuous inference in MaxMSP: jasper-zheng/streamable-stable-audio-open.

• Terrain: stable-audio-terrain-demo.pt
  Config: 2 inputs 64 outputs, range: -8, -8, 8, 8