Kandinsky LabKVAE
KVAE Audio
KVAE Audio compresses and reconstructs waveforms for speech, music and general sound. It provides continuous audio latents for a separate generative model.
Research release
Use the latent representation as a model component
The released package supports waveform reconstruction and latent encoding. Its generative evaluations hold the separate generator fixed, so they do not establish a standalone text-to-audio model.kandinskylab/KVAE-Audio — model card and files
Keep the encoder and decoder paired
Practical guidance. A compressed representation is useful only with a compatible consumer and decoder. Replacing an autoencoder can change the latent distribution even when tensor dimensions match.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | kandinskylab/KVAE-Audio |
| Parameters | 166.9Mkandinskylab/KVAE-Audio — model card and files |
| Audio rate | 48 kHzkandinskylab/KVAE-Audio — model card and files |
| Latent channels | 64kandinskylab/KVAE-Audio — model card and files |
| Weight license | MITkandinskylab/KVAE-Audio — model card and files |