Elamur Research⌕
☰
home / papers / 2026 / Central Kurdish Text-to-Speech and Its Application in Speech-to-Text Translation

Central Kurdish Text-to-Speech and Its Application in Speech-to-Text Translation

Summary

Central Kurdish, spoken by about 8 million people, has very few open speech synthesis resources. Earlier systems were either unreleased or, like Meta's MMS, lacked thorough evaluation and voice cloning. The authors built two new single-speaker datasets, one male and one female, from audiobook podcasts of roughly 11 hours each. They segmented the audio with energy-based VAD, transcribed it with a fine-tuned Seamless ASR and then corrected the transcripts by hand. They also evaluated the studio-recorded Giganet corpus for the first time. Three mono-speaker F5-TTS models were fine-tuned from the English checkpoint, with Asosoft text normalization and G2P conversion to fit the model's vocabulary. Objective metrics and a subjective MOS test indicate that the audiobook-based voices are competitive with the studio-based one. The authors then used the models, with random voice prompts, to generate a fully synthetic corpus for fine-tuning Whisper on Kurdish-to-English speech translation. This shows that cheap web-derived data can bootstrap TTS and downstream speech tools for low-resource languages. The datasets and models are released on Hugging Face under CC BY-NC-ND 4.0.

Key points

  • Datasets: audiobook-M (10h51, 6,044 utterances), audiobook-F (10h54, 5,572) and studio-M/Giganet (13h35, 6,055). Each dataset held out 500 test samples. The audiobook sets come from the biography domain, while studio-M covers multiple domains.
  • Training setup: F5-TTS (flow matching with a diffusion transformer and the Vocos vocoder) was fine-tuned separately for each speaker on a single RTX8000 GPU over about 2 days. The ASR used for transcription reaches roughly 8% WER on the Asosoft benchmark.
  • Evaluation: naturalness was measured with the Squim MOS predictor and the AASIST deepfake score, alongside signal-quality and intelligibility metrics. A human MOS test was used to validate the automatic protocol.
  • Speech translation: Whisper v3 large was fine-tuned on 340k synthetic utterances (514h), using random TTS models and prompts drawn from 5,000 Common Voice clips. BLEU improved from 2.31 to 27.23 on Asosoft and from 1.51 to 18.50 on FLEURS (CKB→ENG).
  • Limitation: ECAPA-TDNN embeddings visualized with T-SNE show that prompting clones speakers only partially, because the training voice still leaks into the synthetic speech.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages

Cited by

1 paper of our library cites it (of the 140673 whose bibliographies we read); each has a page here.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/4hfwowidu34u

@inproceedings{mohammadamini2026central,
  title = {{Central Kurdish Text-to-Speech and Its Application in Speech-to-Text Translation}},
  author = {Mohammadamini, Mohammad and Shamsi, Meysam and Tahon, Marie},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/4hfwowidu34u},
  url = {https://doi.org/10.63317/4hfwowidu34u}
}

References

37 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Hawraz A. Ahmad and Tarik A. Rashid. 2024. Gigant-ktts dataset: Towards building an extensive gigant dataset for kurdish text-to-speech systems. Data in Brief, 55:110753. find
  2. Rosana Ardila, Megan Branson, Kelly Davis, Michael Henretty, Mike Kohler, Josh Meyer, Reuben Morais, Lindsay Saunders, Francis M. Tyers, and Gregor Weber. 2020. Common voice: A massively-multilingual speech corpus. In Proceedings of the 12th Language Resources and Evaluation Conference (LREC 2020),… find our card
  3. Loïc Barrault, Yu-An Chung, Mariano Coria Meglioli, David Dale, Ning Dong, Paul-Ambroise Duquenne, Hady Elsahar, Hongyu Gong, Kevin Heffernan, et. all., and SEAMLESS Communication Team. 2025. Joint speech and text machine translation for up to 100 languages. Nature, 637(8046):587–593. find
  4. Edresson Casanova, Kelly Davis, Eren Gölge, Görkem Göknar, Iulian Gulea, Logan Hart, Aya Aljafari, Joshua Meyer, Reuben Morais, Samuel Olayemi, and Julian Weber. 2024. Xtts: a massively multilingual zero-shot text-to-speech model. In Interspeech 2024, pages 4978–4982. find
  5. Edresson Casanova, Christopher Dossman Shulby, Andrey Korolev, Amaury Cândido Júnior, André S. Soares, Sandra M. Aluísio, and Moacir Antonelli Ponti. 2023. Asr data augmentation in low-resource settings using cross-lingual multi-speaker tts and cross-lingual voice conversion. In Proceedings of Inter… find
  6. Yushen Chen, Zhikang Niu, Ziyang Ma, Keqi Deng, Chunhui Wang, Jian Zhao, Kai Yu, and Xie Chen. 2025. F5-TTS: A fairytaler that fakes fluent and faithful speech with flow matching. find
  7. Alexis Conneau, Min Ma, Simran Khanuja, Yu Zhang, Vera Axelrod, Siddharth Dalmia, Jason Riesa, Clara Rivera, and Ankur Bapna. 2023. Fleurs: Few-shot learning evaluation of universal representations of speech. In 2022 IEEE Spoken Language Technology Workshop (SLT), pages 798–805. find our card
  8. Brecht Desplanques, Jenthe Thienpondt, and Kris Demuynck. 2020. Ecapa-tdnn: Emphasized channel attention, propagation and aggregation in tdnn based speaker verification. In Interspeech 2020, pages 3830–3834. find
  9. David Guennec, Lily Wadoux, Aghilas Sini, Nelly Barbot, and Damien Lolive. 2023. Voice cloning: Training speaker selection with limited multi-speaker corpus. In 12th ISCA Speech Synthesis Workshop, SSW 2023, Grenoble, France, August 26-28, 2023, pages 170–176. ISCA. find
  10. Wen-Chin Huang, Erica Cooper, Yu Tsao, Hsin-Min Wang, Tomoki Toda, and Junichi Yamagishi. 2022. find
  11. The VoiceMOS Challenge 2022. In Interspeech 2022, pages 4536–4540. find
  12. Jee-weon Jung, Hee-Soo Heo, Hemlata Tak, Hyejin Shim, Joon Son Chung, Bong-Jin Lee, Ha-Jin Yu, and Nicholas Evans. 2022. Aasist: Audio anti-spoofing using integrated spectro-temporal graph attention networks. In ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Process… find
  13. Chanwoo Kim and Richard M. Stern. 2008. Robust signal-to-noise ratio estimation based on waveform amplitude distribution analysis. In Interspeech 2008, pages 2598–2601. find
  14. Jaehyeon Kim, Jungil Kong, and Juhee Son. 2021. Conditional variational autoencoder with adversarial learning for end-to-end text-to-speech. In Proceedings of the 38th International Conference on Machine Learning, volume 139 of Proceedings of Machine Learning Research, pages 5530– 5540. PMLR. find
  15. Zhaolin Li, Yining Liu, Danni Liu, Tuan Nam Nguyen, Enes Yavuz Ugan, Tu Anh Dinh, Carlos Mullov, Alexander Waibel, and Jan Niehues. 2025. KIT’s low-resource speech translation systems for IWSLT2025: System enhancement with synthetic data and model regularization. In Proceedings of the 22nd Internati… find
  16. Yunpeng Liu, Xukui Yang, and Dan Qu. 2024. Exploration of whisper fine-tuning strategies for low-resource asr. EURASIP Journal on Audio, Speech, and Music Processing, 2024(1):29. find
  17. Johannes Abraham Louw. 2023. Cross-lingual transfer using phonological features for resource-scarce text-to-speech. In 12th ISCA Speech Synthesis Workshop (SSW2023), pages 55–61. find
  18. Florian Lux, Julia Koch, Sarina Meyer, Thomas Bott, Nadja Schauffler, Pavel Denisov, Antje Schweitzer, and Ngoc Thang Vu. 2023. The ims toucan system for the blizzard challenge 2023. In 18th Blizzard Challenge Workshop, pages 40– 45. find
  19. Aso Mahmudi and Hadi Veisi. 2021. Automated grapheme-to-phoneme conversion for central kurdish based on optimality theory. Computer Speech & Language, 70:101222. find
  20. Aso Mahmudi, Hadi Veisi, Mohammad MohammadAmini, and Hawre Hosseini. 2019. Automated kurdish text normalization. find
  21. Pauline Mas, Natacha Miniconi, Kevin Vythelingum, Meysam Shamsi, Aghilas Sini, Lu Zuo, Elias Okat, and Marie Tahon. 2025. VO2Lium : Voxy-gen and LIUM contribution for Blizzard 2025. In The Blizzard Challenge 2025, pages 18–23. find
  22. Natacha Miniconi, Meysam Shamsi, Aghilas Sini, and Anthony Larcher. 2025. Using a deepfake classifier to rank speech synthesis quality. In The Speech Synthesis Workshop - ISCA. find
  23. Tomoya Mizumoto, Atsushi Kojima, Yusuke Fujita, Lianbo Liu, and Yui Sudo. 2025. Is Synthetic Data Truly Effective for Training Speech Language Models? In Interspeech 2025, pages 1808–1812. find
  24. Mohammad Mohammadamini, Daban Jaff, Sara Jamal, Ibrahim Ahmed, Hawkar Omar, Darya Sabr, Marie Tahon, and Antoine Laurent. 2025a. Kuvost: A large-scale human-annotated English to Central Kurdish speech translation dataset driven from English common voice. In Proceedings of the 22nd International Conf… find
  25. Mohammad Mohammadamini, Aghilas Sini, Marie Tahon, and Antoine Laurent. 2025b. Scaling pseudo-labeling data for end-to-end low-resource speech translation (the case of Kurdish language). In Interspeech 2025, pages 898–902. find
  26. Sabat Salih Muhamad, Hadi Veisi, Aso Mahmudi, Abdulhady Abas Abdullah, and Farhad Rahimi. 2024. Kurdish end-to-end speech synthesis using deep neural networks. Natural Language Processing Journal, 8:100096. find
  27. Alex Peiró-Lilja, Guillermo Cámbara, Mireia Farrús, and Jordi Luque. 2022. Naturalness and Intelligibility Monitoring for Text-to-Speech Evaluation. Proc. Speech Prosody, pages 445–449. find
  28. Vineel Pratap, Andros Tjandra, Bowen Shi, Paden Tomasello, Arun Babu, Sayani Kundu, Ali Elkahky, Zhaoheng Ni, Apoorv Vyas, Maryam Fazel-Zarandi, Alexei Baevski, Yossi Adi, Xiaohui Zhang, Wei-Ning Hsu, Alexis Conneau, and Michael Auli. 2024. Scaling speech technology to 1,000+ languages. Journal of M… find our card
  29. Yu Pu, Xiaoqian Liu, Guangyu Zhang, Zheng Yan, Wei-Qiang Zhang, and Xie Chen. 2025. Empowering Large Language Models for End-to-End Speech Translation Leveraging Synthetic Data. In Interspeech 2025, pages 26–30. find
  30. Mirco Ravanelli, Titouan Parcollet, Peter Plantinga, Aku Rouhe, Samuele Cornell, Loren Lugosch, Cem Subakan, Nauman Dawalatabad, Abdelwahab Heba, Jianyuan Zhong, Ju-Chieh Chou, Sung-Lin Yeh, Szu-Wei Fu, Chien-Feng Liao, Elena Rastorgueva, François Grondin, William Aris, Hwidong Na, Yan Gao, Renato D… — [arXiv:2106.04624](https://arxiv.org/abs/2106.04624) find
  31. Antony W Rix, John G Beerends, Michael P Hollier, and Andries P Hekstra. 2001. Perceptual evaluation of speech quality (pesq)-a new method for speech quality assessment of telephone networks and codecs. In 2001 IEEE international conference on acoustics, speech, and signal processing. Proceedings (C… find
  32. Jaffer Sheyholislami. 2015. The Kurds: History, Religion, Language, Politics, chapter Language Varieties of the Kurds. Austrian Federal Ministry of the Interior. find
  33. Jaffer Sheyholislami. 2021. The Cambridge History of the Kurds, chapter The History and Development of Literary Central Kurdish. Cambridge University Press. find
  34. Hubert Siuzdak. 2023. Vocos: Closing the gap between time-domain and fourier-based neural vocoders for high-quality audio synthesis. arXiv preprint arXiv:2306.00814. — [arXiv:2306.00814](https://arxiv.org/abs/2306.00814) find
  35. Marie Tahon, Gwénolé Lecorvé, Damien Lolive, and Raheel Qader. 2017. Perception of expressivity in TTS: linguistics, phonetics or prosody? In Statistical Language and Speech Processing (SLSP), volume 10583, page 262. find
  36. Hadi Veisi, Hawre Hosseini, Mohammad MohammadAmini, Wirya Fathy, and Aso Mahmudi. 2022. Jira: a central kurdish speech recognition system, designing and building speech corpus and pronunciation lexicon. Lang. Resour. Eval., 56(3):917–941. find
  37. Xin Wang, Héctor Delgado, Hemlata Tak, Jee weon Jung, Hye jin Shim, Massimiliano Todisco, Ivan Kukanov, Xuechen Liu, Md Sahidullah, Tomi H. Kinnunen, Nicholas Evans, Kong Aik Lee, and Junichi Yamagishi. 2024. Asvspoof 5: crowdsourced speech data, deepfakes, and adversarial attacks at scale. In The A… find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/mohammadamini-2026-central-kurdish-text-speech-its/