cs.CLMar 8, 2026

Bolbosh: Script-Aware Flow Matching for Kashmiri Text-to-Speech

Authors: Tajamul Ashraf, Burhaan Rasheed Zargar, Saeed Abdul Muizz, Ifrah Mushtaq, Nazima Mehdi, Iqra Altaf Gillani, Aadil Amin Kak, Janibul Bashir

Organizations: King Abdullah University of Science and Technology (KAUST) · Gaash Lab, National Institute of Technology Srinagar · Department of Linguistics, University of Kashmir.

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Kashmiri is spoken by around 7 million people but remains critically underserved in speech technology, despite its official status and rich linguistic heritage. The lack of robust Text-to-Speech (TTS) systems limits digital accessibility and inclusive human-computer interaction for native speakers. In this work, we present the first dedicated open-source neural TTS system designed for Kashmiri. We show that zero-shot multilingual baselines trained for Indic languages fail to produce intelligible speech, achieving a Mean Opinion Score (MOS) of only 1.86, largely due to inadequate modeling of Perso-Arabic diacritics and language-specific phonotactics. To address these limitations, we propose Bolbosh, a supervised cross-lingual adaptation strategy based on Optimal Transport Conditional Flow Matching (OT-CFM) within the Matcha-TTS framework. This enables stable alignment under limited paired data. We further introduce a three-stage acoustic enhancement pipeline consisting of dereverberation, silence trimming, and loudness normalization to unify heterogeneous speech sources and stabilize alignment learning. The model vocabulary is expanded to explicitly encode Kashmiri graphemes, preserving fine-grained vowel distinctions. Our system achieves a MOS of 3.63 and a Mel-Cepstral Distortion (MCD) of 3.73, substantially outperforming multilingual baselines and establishing a new benchmark for Kashmiri speech synthesis. Our results demonstrate that script-aware and supervised flow-based adaptation are critical for low-resource TTS in diacritic-sensitive languages. Code and data are available at: https://github.com/gaash-lab/Bolbosh.

Figures & tables

SplitRASAIndicVoices-RTotal Duration
Train25.49 h (15,898)43.61 h (17,284)69.10 h
Validation7.23 h (4,542)0.00 h (0)7.23 h
Test3.56 h (2,272)0.00 h (0)3.56 h
Total36.28 h43.61 h79.89 h
Table 1: Duration and utterance statistics for the Kashmiri TTS dataset. IndicVoices-R was used exclusively during training for acoustic prior modeling.
Model FamilyModelConditionWER
IndicConformer [ 28 ]RNN-TWith Diacritics66.59
(AI4Bharat)RNN-TNo Diacritics41.20
 CTCNo Diacritics44.00
OmniASR [ 29 ]CTC (300M)With Diacritics94.34
(Meta)CTC (300M)No Diacritics90.06
 LLM (7B)No Diacritics87.67
Table 2: Benchmarking ASR Models for Kashmiri Proxy Evaluation. We compare WER across architectures and normalization strategies.
ModelConditionMCDrWER (%)WER
BolboshWith Diacritics3.734.140.6935
 No Diacritics–13.230.4665
IndicParlerWith Diacritics4.7346.750.9772
 No Diacritics–100.320.8253
Table 3: TTS Benchmark Results. MCD, rWER, and WER ( ↓\downarrow ) lower is better.
SystemMOS ( ↑\uparrow )95% CI
Human (Ground Truth)4.614±\pm 0.059
Bolbosh (Ours)3.634±\pm 0.061
IndicParler (Baseline)1.864±\pm 0.065
Table 4: Mean Opinion Score (MOS) Results for Kashmiri TTS Systems with 95% Confidence Intervals.
Figure 1: Mel-spectrogram comparison of 2 Kashmiri utterances. The utterances were synthesized for the following text: Right (IPA): [ \textipa me ru: \textsubbridge d n1 pA:n2s \textsubbridge tA:m jel1 lUkow \textsubbridge t2 \textsubbridge t \super h @n \textsubbridge d2s p2kn1 bA:p@ \textsubbridge t \super h s2 \textrtaild kI pj2 \textrtailt \super h træp \super hIk \textdyoghlig A:m@s m2nz bA:kIjow k \super ho \textsubbridge t1 brõh nj@:rn1 k \super h@: \textsubbridge tr1 g@: \super j \textrtaild \textteshlig 2l@:w j2 \textsubbridge t \super h pj2 \textrtailt \super h b1 p@kA:n o:sUs] Left (IPA): [ \textipa me \textteshlig \super h1 k \super hoSi: zI seh2 \textsubbridge t m2rk@z@s m2nz j1m h2p \super h \textsubbridge t1 kIs @:k \super hr@s pj2 \textrtailt \super h m2Sw@r1 k \super h@: \textsubbridge tr1 wA:rj@h \textsubbridge t2 \textdyoghlig rUb1 kA:r \textrtaild A:k \super h \textrtailt @r \textsubbridge d2s \textsubbridge tIjA:b]
More in this topic

TTS Synthesis

Text-to-speech synthesis converts written language into intelligible, natural-sounding speech with control over voice, prosody, style, and timing.

See all 275 papers

Explore similar work

CardsList
May 4, 2026

Tibetan-TTS:Low-Resource Tibetan Speech Synthesis with Large Model Adaptation

Tibetan text-to-speech (TTS) has long been challenged by scarce speech resources, significant dialectal variation, and the complex mapping between written text and spoken pronunciation. To address these issues, this work presents, to the best of our knowledge, the first large-model-based Tibetan TTS system in the industry, built upon a large speech synthesis model developed by Xingchen AGI Lab. The proposed system integrates data quality enhancement, Tibetan-oriented text representation and tokenizer adaptation, and cross-lingual adaptive training for low-resource Tibetan speech synthesis. Experimental results show that the system can generate stable, natural, and intelligible Tibetan speech under low-resource conditions. In subjective evaluation, the MOS scores of the syllable-level and BPE-based systems reach 4.28 and 4.35, while their pronunciation accuracies reach 97.6% and 96.6%, respectively, outperforming an external commercial Tibetan TTS interface. These results demonstrate that combining a large-model backbone with Tibetan-oriented text representation adaptation and cross-lingual adaptive training enables highly usable low-resource Tibetan speech synthesis, and also provides a technical foundation for future unified multi-dialect Tibetan speech synthesis.

Jiaxu He, Chao Wang, Jie Lian +4

Xingchen AGI Lab, China Telecom Artificial Intelligence Technology Co. Ltd · Qinghai Normal University, Xining, China · University of Electronic Science and Technology of China,Chengdu,China +1

TTS SynthesisLow-Resource TTS SynthesisLow-Resource Language Processing
AskPDF
Jun 14, 2026

Koshur Diacritizer: A Byte-Level Sequence-to-Sequence Model for Kashmiri Diacritic Restoration

Kashmiri, an Indo-Aryan language written in a modified Perso-Arabic script, frequently omits diacritic marks in digital text, creating ambiguity and challenging downstream NLP applications. We present Koshur Diacritizer, a ByT5-small byte-level sequence-to-sequence model for restoring diacritics in Kashmiri text. To support this task, we release a publicly available dataset of 23.7k aligned undiacritized diacritized Kashmiri sentence pairs. The proposed framework combines script-aware normalization, alignment validation, and skeleton-preserving inference to ensure reliable restoration while maintaining the original base-letter sequence. Experimental results on a held-out test set achieve a DERm of 0.2012 and a WER of 0.2159. Additionally, evaluation by a native Kashmiri linguistic expert yields a mean accuracy of 77.5%. The dataset, model, and source code are publicly released to provide a reproducible baseline for Kashmiri diacritic restoration and future low-resource language research.

Haq Nawaz Malik, Nahfid Nissar, Faizan Iqbal

ORCID: 0009-0003-1994-7640 · ORCID: 0009-0002-2805-4687 · ORCID: 0009-0002-8998-9347

Low-Resource Language Processing
AskPDF
Apr 13, 2026

Ti-Audio: The First Multi-Dialectal End-to-End Speech LLM for Tibetan

Recent advances in Speech Large Language Models (Speech-LLMs) have made significant progress, greatly enhancing multimodal interaction capabilities.However, their application in low-resource and dialect-diverse environments still faces challenges. The severe scarcity of Tibetan data, coupled with the phonetic differences among its major dialects (Ü-Tsang, Amdo, and Kham), is a prime example of this challenge. This paper proposes Ti-Audio, the first multi-dialectal end-to-end Speech-LLM for Tibetan. To efficiently align speech and text, we introduce a Dynamic Q-Former Adapter that extracts essential acoustic features from variable-length speech, ensuring stable cross-modal alignment even with limited data. At the data level, we leverage mutual assistance among related dialects to alleviate data scarcity and employ a temperature-based sampling strategy to maximize this synergy. Experimental results demonstrate that Ti-Audio achieves state-of-the-art performance on Tibetan benchmarks for automatic speech recognition and speech translation. Our work validates the effectiveness of cross-dialectal cooperation and provides a scalable paradigm for the development of Speech-LLM in low-resource scenarios.

Jialing Wang, Yue Zhao, Yuhao Zhang +5

Minzu University of China, Beijing, China · The Chinese University of Hong Kong, Shenzhen, China

Speech Language ModelsLow-Resource Speech RecognitionAutomatic Speech RecognitionSpeech Translation
AskPDF