EnCodec · 32 kHz
Represent music as audio tokens, including the codec used by MusicGen.
Audio codecsOpen-source code
- Start here
- Use it when an existing model expects these exact audio tokens. A codec cannot turn a description into music without a separate generative model.
- Training data
- The linked 32 kHz music checkpoint is part of AudioCraft; do not assume the 24/48 kHz EnCodec training setup applies unchanged.
- Training rights
- A full source-recording ledger is not supplied in the linked card. AudioCraft training music is not redistributed with the model.
- License & output terms
- Code: MIT. Linked 32 kHz weights: CC BY-NC 4.0.
- Limitations
- Checkpoint-specific rates and codebooks; lossy reconstruction and non-commercial weights. Tokens represent sound, not MIDI notes.
- Technique
- Neural audio compression / RVQ
- Runtime
- Local inference; encoder and decoder must use matching configurations.
- 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.