Descript Audio Codec
Encode audio into discrete tokens and reconstruct it for compression or generative-model research.
Audio codecsOpen-source code
- Start here
- Encode and decode a reference clip first. Listen for lost attacks, stereo detail and high-frequency texture before building a generator around its tokens.
- Training data
- A mixture of speech, music and environmental audio described in the paper. The README does not provide a file-level rights ledger.
- Training rights
- The released weights do not redistribute or clear all source recordings; inspect the paper's data mixture for your use case.
- License & output terms
- Code and released weights: MIT, as stated by the repository.
- Limitations
- Lossy reconstruction. A codec supplies an audio representation, not text-to-music generation by itself.
- Technique
- Residual vector quantization / autoencoder
- Runtime
- Local encoding/decoding; select the matching 16, 24 or 44.1 kHz model.
- 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.