Generative pre-training with flow matching in speech technology shows promising results for various downstream tasks.
