How Much Audio Is Left In An Embedding? An Inversion Audit Of Audio Encoders
Using a shared Stable Audio Open latent diffusion decoder, we reconstruct five-second, 44.1-kHz stereo music from frozen representations produced by supervised classifiers (VGGish, ConvNeXt), an audio-text contrastive model (CLAP), and a waveform-reconstruction model (EnCodec).
ProofPaper ↗
Key points
- Pretrained audio encoders are reused for downstream tasks that are often unknown when the encoder is trained, so their usefulness depends partly on which signal properties survive the pretext objective.
- We study this retained information through paired source reconstruction.
- Evaluating on the Million Song Dataset (MSD), we find clear differences in reconstructability across encoder families, while within encoder comparisons show improved recovery when finer temporal or spectral structure is exposed.
- Even compressed task oriented embeddings support reconstructions that preserve measurable source specificity and high level musical content.
Sources (1)
- [1]How Much Audio Is Left In An Embedding? An Inversion Audit Of Audio EncodersarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 04:24 PM
Using a shared Stable Audio Open latent diffusion decoder, we reconstruct five-second, 44.1-kHz stereo music from frozen representations produced by supervised classifiers (VGGish, ConvNeXt), an audio-text contrastive model (CLAP), and a waveform-reconstruction model (EnCodec).
Pretrained audio encoders are reused for downstream tasks that are often unknown when the encoder is trained, so their usefulness depends partly on which signal properties survive the pretext objective.
Extractive summary: sentences quoted from the sources.