This paper introduces a discrete diffusion model (DDM) framework for text-aligned speech tokenization and reconstruction. By replacing the auto-regressive speech decoder with a discrete diffusion counterpart, our model achieves significantly better reconstruction quality, stronger ASR performance, and faster inference. We provide a comprehensive analysis of applying DDMs to speech reconstruction, examining sampler choices, inference steps, and robustness to length-scale estimation errors. Furthermore, we improve the original TASTE by systematically comparing vector quantization modules, showing that FSQ yields up to a 35% relative WER reduction and +0.14 UT-MOS improvement over RVQ for AR models, while also enhancing DDM performance. Our model generates speech in just 10 denoising steps and even supports single-step generation with only minor quality degradation.
@article{arxiv.2509.20060,
title = {Discrete Diffusion for Generative Modeling of Text-Aligned Speech Tokens},
author = {Pin-Jui Ku and He Huang and Jean-Marie Lemercier and Subham Sekhar Sahoo and Zhehuai Chen and Ante Jukić},
journal= {arXiv preprint arXiv:2509.20060},
year = {2025}
}