#natural language process...
May 2026
MELD: Mel-Spectrogram-Based Speech Language Modeling with Discrete Latent Variables
A discrete latent variable model on mel spectrograms that jointly optimizes the encoder and the speech language model is introduced that effectively alleviates common issues in autoregressive mel spectrogram modeling.
Sung-Lin Yeh, Wei Zhou, Gil Keren et al.
· arXiv.org · 1 citation