UR-BERT: New Text Encoder Scales Text-to-Speech to 495 Languages Using Universal Romanization
Researchers have proposed UR-BERT, a text encoder for text-to-speech (TTS) systems that scales to 495 languages by converting diverse writing systems into a shared Romanized representation. Conventional grapheme-to-phoneme approaches are constrained to roughly 100 languages due to limited linguistic resources, a gap UR-BERT addresses through universal Romanization combined with a speech token prediction training objective. The work, accepted to Interspeech 2026, significantly broadens the reach of high-quality speech synthesis to low-resource and previously unsupported languages.
UR-BERT is a newly proposed text encoder designed to overcome a longstanding bottleneck in multilingual text-to-speech synthesis: the scarcity of reliable grapheme-to-phoneme (G2P) resources beyond roughly 100 well-resourced languages. By converting diverse writing systems into a unified Romanized transcription, the model extends coverage to 495 languages within a single shared representation space. To improve phonetic accuracy and alignment between text and speech, the authors introduce a speech token prediction objective during training, which guides the encoder toward speech-aware phonetic representations without requiring large amounts of data. Experiments demonstrate that TTS systems built on UR-BERT consistently outperform recent text encoder baselines across a wide range of languages and resource conditions. The model also shows strong generalization to languages not seen during training, suggesting robustness beyond its training distribution. The paper has been accepted to Interspeech 2026, and code is publicly available on GitHub.
What's missing
The paper does not detail the specific quality or source of Romanization mappings for all 495 languages, which could affect performance for languages with irregular or contested romanization conventions. It is also unclear how subjective speech quality (e.g., naturalness as rated by native speakers) was evaluated across low-resource languages, and whether the 495-language coverage includes languages with very limited training data or relies on zero-shot transfer. The degree to which Romanization introduces ambiguity or information loss for tonal or morphologically complex languages is not addressed in the abstract.
What different sources said
- arXiv cs.CLCenter
UR-BERT: Scaling Text Encoders for Massively Multilingual TTS Through Universal Romanization and Speech Token Prediction
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.