AI & ML Open Release

Delivers a state-of-the-art universal phone recognition model across 100+ languages with full open-source release.

April 2, 2026

Original Paper

An Empirical Recipe for Universal Phone Recognition

Shikhar Bharadwaj, Chin-Jou Li, Kwanghee Choi, Eunjung Yeo, William Chen, Shinji Watanabe, David R. Mortensen

arXiv · 2603.29042

The Takeaway

PhoneticXEUS bridges the gap between high-resource English models and low-resource multilingual needs. By establishing an empirical recipe for phonetic recognition that generalizes across language families and accents, it provides a foundational tool for global speech-to-text and linguistic analysis that was previously gated by data scale and compute.

From the abstract

Phone recognition (PR) is a key enabler of multilingual and low-resource speech processing tasks, yet robust performance remains elusive. Highly performant English-focused models do not generalize across languages, while multilingual models underutilize pretrained representations. It also remains unclear how data scale, architecture, and training objective contribute to multilingual PR. We present PhoneticXEUS -- trained on large-scale multilingual data and achieving state-of-the-art performance