Pre-trained RAVE streaming models are available for immediate use without training

ACIDS-IRCAM hosts a collection of pre-trained RAVE streaming models available for free download. These are already exported as .ts files with streaming mode enabled, ready to load directly into nn~ or the RAVE VST. Using pre-trained models is the fastest entry point to RAVE-based live performance: no training infrastructure needed. The models represent different timbral corpora; darbouka (percussion) is commonly used in demonstration patches and tutorials. Note the pre-trained model packs are CC-BY-NC (NonCommercial).

Examples

Download a _streaming.ts model → place it in the Max file paths → nn~ darbouka forward 2048 in a patch → feed audio in for immediate timbre transfer.

Assessment

Name two scenarios where a performer would prefer to train their own RAVE model rather than use a pre-trained one, and two scenarios where the pre-trained models are sufficient.