PAST: Phonetic-Acoustic Speech Tokenizer
This addresses the problem of efficient and accurate speech tokenization for real-time applications, though it appears incremental by building on existing tokenization methods with supervised phonetic integration.
The paper tackles speech tokenization by introducing PAST, a framework that jointly models phonetic information and signal reconstruction without external pretrained models, achieving superior performance in metrics like phonetic representation and speech reconstruction, and also enhances speech language models.
We present PAST, a novel end-to-end framework that jointly models phonetic information alongside signal reconstruction, eliminating the need for external pretrained models. Unlike previous approaches that rely on pretrained self-supervised models, PAST employs supervised phonetic data, directly integrating domain knowledge into the tokenization process via auxiliary tasks. Additionally, we introduce a streamable, causal variant of PAST, enabling real-time speech applications. Results demonstrate that PAST surpasses existing evaluated baseline tokenizers across common evaluation metrics, including phonetic representation and speech reconstruction. Notably, PAST also achieves superior performance when serving as a speech representation for speech language models, further highlighting its effectiveness as a foundation for spoken language generation. To foster further research, we release the full implementation. For code, model checkpoints, and samples see: https://pages.cs.huji.ac.il/adiyoss-lab/PAST