RAVE
Learn a compact timbre space and explore sound transformations as a playable instrument.
Neural instrumentsOpen weights
- Maru integration
Maru extension available
Runs as a Maru browser module. Supported checkpoints and host requirements are documented on the extension page.
- Start here
- Train on one coherent sound family. Explore whether it makes a responsive instrument before expanding the training collection.
- Training data
- Train on a chosen collection of recordings. Each published RAVE checkpoint has its own sound corpus.
- Training rights
- Use recordings you own or have permission to train on; third-party checkpoint provenance must be checked individually.
- License & output terms
- Linked repository: CC BY-NC 4.0. Checkpoint and host/plugin terms are separate.
- Limitations
- Non-commercial repository terms. A model learns the range of its corpus; timbre transfer does not preserve every transient, articulation or pitch equally.
- Technique
- Variational autoencoder
- Runtime
- Designed for fast local inference after training; training is a separate GPU workload.
- Parameters
- Not itemized for this release/family. Parameter sizes shown elsewhere may exclude encoders, decoders or other runtime components.
Sources reviewed . Training and rights notes summarize provider disclosures. A model license does not clear every source recording or generated output. Parameter counts refer to the named variant or component, not total runtime memory.