kaldi-cy 2610, released October 2026 Copyright 2026 Bangor University Developed by the Language Technologies Unit (Uned Technolegau Iaith), Bangor University. Licensed under the Apache License, Version 2.0 (the "License"). You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 NAMES This licence grants no permission to use the names "Bangor University", "Prifysgol Bangor", "Language Technologies Unit", "Uned Technolegau Iaith", "techiaith", the University's logos, or the names of its products or staff, except to describe the origin of the model as this NOTICE does. WHAT THIS MODEL IS This is a research model, trained on the data and evaluated by the methods described in its model card. It is not a certified product and is not a substitute for a qualified translator, transcriber, clinician or other professional. AUTOMATIC PROCESSING MAKES MISTAKES Output from this model, whether a translation, a transcript, generated text or speech, or any other result, can be wrong while appearing fluent and confident. Known failure modes include omitted or inserted words, incorrect names, numbers, dates and negations, and reduced accuracy on dialects, noisy audio or subjects outside the training data. Prifysgol Bangor University therefore advises that output is treated as provisional and is checked and corrected by a competent person before it is relied on as a record, used in any decision affecting a person, or shared or published. Where the model is used interactively or in real time (for example live captions, dictation, suggestions in an editor or read-aloud), users should be told that what they see or hear is automatic and may be wrong. The University asks anyone who builds on or redistributes this model to pass this warning on to their own users. NOT FIT FOR The model has not been evaluated for, and should not be used in, medical, legal, financial, safety-critical or emergency settings, or in any decision affecting a person's rights, health, liberty or livelihood, unless a qualified person reviews every output. USE IN WALES: EQUAL TREATMENT OF WELSH AND ENGLISH This model was developed to help organisations provide services in Welsh that are as good as those they provide in English, not to replace Welsh-language provision. Under section 1 of the Welsh Language (Wales) Measure 2011 (2011 nawm 1), Welsh has official status in Wales, and that status is given legal effect by, among other things, the treatment of Welsh no less favourably than English (s. 1(2)(b)) and by standards of conduct relating to the use of Welsh (s. 1(3)(e)). These standards are made and enforced under Part 4 of the Measure. Bodies still operating a Welsh language scheme under Part II of the Welsh Language Act 1993 (c. 38) must give effect to the principle in section 5 of that Act that, in the conduct of public business and the administration of justice in Wales, Welsh and English should be treated on a basis of equality. Prifysgol Bangor University asks anyone deploying this model in Wales to consider whether the intended use meets those requirements. Offering automated Welsh where English is served by people, publishing machine translation into Welsh without the review given to English text, or otherwise using this technology to replace rather than support Welsh-language provision is contrary to the purpose for which the model was released. Which duties apply depends on the body and the regulations or scheme that cover it, and responsibility for compliance rests with the body deploying the model. See the Welsh Language Commissioner's regulatory policy statement on artificial intelligence and the Welsh language, Technology, Websites and Software: Welsh Language Considerations, and the advice document Bilingual drafting, translation and using Welsh face to face, all linked from the model card. The legislation is at legislation.gov.uk/mwa/2011/1 and legislation.gov.uk/ukpga/1993/38. FURTHER INFORMATION AND ERROR REPORTS Training data, evaluation results and known limitations: https://huggingface.co/techiaith/kaldi-cy Report errors or concerns to: techiaith@bangor.ac.uk PRONUNCIATION DICTIONARY The decoding graph (HCLG.fst) and word list (words.txt) in this bundle were built with pronunciations from Geiriadur Ynganu Bangor, the Bangor Pronouncing Dictionary, https://github.com/techiaith/geiriadur-ynganu-bangor, redistributed here under its own licence, which requires the following to be kept with any binary redistribution: Copyright (C) 2025 Prifysgol Bangor University. All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. The contents of this file are deemed to be source code. 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. THIS SOFTWARE IS PROVIDED BY PRIFYSGOL BANGOR UNIVERSITY ``AS IS'' AND ANY EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL PRIFYSGOL BANGOR UNIVERSITY NOR ITS EMPLOYEES BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. BUILT WITH Trained with Kaldi (Apache License 2.0) and pocolm (Apache License 2.0), and packaged for Vosk (Apache License 2.0), following the vosk-cymraeg recipe by Gweltaz Duval-Guennoc, Preben Vangberg and Sasha Wanasky (MIT License, Cymru-Breizh Agile Cymru Project, https://github.com/Cymru-Breizh-Agile-Cymru-Project/vosk-cymraeg). None of their code is included in this bundle. TRAINING DATA ATTRIBUTIONS The Apache License applies to the model weights, code and documentation. It does not license the data the model was trained on, which is not distributed with the model and remains under the terms of its original providers, acknowledged here as those terms require. Dedicated to the public domain under CC0 1.0 Universal (https://creativecommons.org/publicdomain/zero/1.0/). CC0 asks for no attribution; the sources are listed so that the training data can be traced to its datasheet. Banc Trawsgrifiadau Bangor (CC0 1.0), collected by Language Technologies Unit, Bangor University. Collection process, consent terms and contact: https://huggingface.co/datasets/techiaith/banc-trawsgrifiadau-bangor Corpws CLlLC / WLGA (CC0 1.0), collected by Language Technologies Unit, Bangor University. Collection process, consent terms and contact: https://huggingface.co/datasets/techiaith/corpws-clllc-wlga Mozilla Common Voice 23.0, Welsh (CC0 1.0), collected by Mozilla Foundation and the volunteers who contributed their voices. Collection process, consent terms and contact: https://commonvoice.mozilla.org/cy/datasets. Obtained from Mozilla under their terms for Common Voice data and held in a private working copy; it is not redistributed with or by this model. Lleisiau ARFOR (CC0 1.0), collected by Cymen, with the Language Technologies Unit, Bangor University. Collection process, consent terms and contact: https://huggingface.co/datasets/cymen-arfor/lleisiau-arfor Mozilla Common Voice, Welsh, re-segmented with voice activity detection (CC0 1.0), collected by Mozilla Foundation and the volunteers who contributed their voices. Collection process, consent terms and contact: https://commonvoice.mozilla.org/cy/datasets. Derived from Common Voice data obtained from Mozilla under their terms and held in a private working copy; it is not redistributed with or by this model.