Techiaith and Dewi Bryn Jones catalogue
A snapshot of all public models, datasets, Hugging Face Spaces demos and collections in the two accounts, organised by use.

Snapshot taken on 5 August 2026. The account Techiaith had 33 models and 28 datasets, while DewiBrynJones had 14 models and 15 datasets. Account contents may change; repository names are retained here as they appear on Hugging Face to avoid ambiguity.
Voice and speech
This section covers speech recognition, Welsh–English transcription, text-to-speech, punctuation and the audio data used in this work. Try the simple browser demo on the Voice page.
Techiaith models — voice and speech (24)
- techiaith/whisper-large-ft-cy-en-ct2
- techiaith/whisper-large-ft-cy-en
- techiaith/cy_en_GB-bu_ttsWelsh–English text-to-speech
- techiaith/kaldi-cy-2601
- techiaith/orpheus3b-cy-en
- techiaith/cy_GB-bu_ttsWelsh text-to-speech
- techiaith/kaldi-cy
- techiaith/microsoft_speecht5_finetuned_bu_tts_cy_en
- techiaith/whisper-large-v3-ft-verbatim-cy-en-ct2
- techiaith/whisper-large-v3-ft-verbatim-cy-en
- techiaith/wav2vec2-base-cy
- techiaith/whisper-base-ft-verbatim-cy-en-cpp
- techiaith/whisper-base-ft-commonvoice-cy-cpp
- techiaith/whisper-large-v3-turbo-ft-verbatim-cy-en
- techiaith/whisper-base-ft-commonvoice-cy-en-cpp
- techiaith/wav2vec2-btb-cv-ft-cv-cy
- techiaith/wav2vec2-xlsr-53-ft-btb-cv-cy
- techiaith/whisper-large-v3-ft-commonvoice-cy
- techiaith/whisper-large-v3-ft-commonvoice-cy-en
- techiaith/fullstop-welsh-punctuation-predictionPunctuation prediction
- techiaith/whisper-tiny-ft-cy-en
- techiaith/wav2vec-xlsr-ft-cy-verbatim
- techiaith/wav2vec2-xlsr-53-ft-cy-en-withlm
- techiaith/wav2vec2-xlsr-53-ft-cy-en
DewiBrynJones models — voice and speech (11)
- DewiBrynJones/whisper-large-v2-ft-cy-2607
- DewiBrynJones/kaldi-cy-2606
- DewiBrynJones/whisper-large-v2-ft-cy-2606
- DewiBrynJones/whisper-large-v2-ft-cy-2603
- DewiBrynJones/whisper-large-v2-ft-cy-2602
- DewiBrynJones/whisper-large-v2-ft-cy-2601
- DewiBrynJones/wav2vec2-xlsr-53-ft-cy-2601
- DewiBrynJones/whisper-large-v2-ft-btb-cv-cvad-ca-wlga-cy-mlx-2511
- DewiBrynJones/whisper-large-v2-ft-btb-cv-cvad-ca-wlga-cy-ct2-2511
- DewiBrynJones/whisper-large-v2-ft-btb-cv-cvad-cven-cvmt-wlga-ca-2512
- DewiBrynJones/whisper-large-v2-ft-btb-cv-cvad-ca-wlga-cy-2511
Techiaith datasets — voice and speech (12)
- techiaith/banc-trawsgrifiadau-bangor
- techiaith/evals-speech-recognition-cy-en
- techiaith/corpws-clllc-wlga
- techiaith/commonvoice_22_0_en__GB_IE
- techiaith/bu-tts-cy-en
- techiaith/eval-kaldi-cy-2503
- techiaith/commonvoice_18_0_cy_en
- techiaith/commonvoice_tts_cy
- techiaith/commonvoice_vad_cy
- techiaith/commonvoice_18_0_cy
- techiaith/commonvoice_16_1_en_cy
- techiaith/commonvoice_16_1_cy
DewiBrynJones datasets — voice and speech (13)
- DewiBrynJones/evals-speech-recognition-cy-en-2606
- DewiBrynJones/preprocessed-whisper-btb-cv-cvad-wlga-ca-2607
- DewiBrynJones/preprocessed-whisper-btb-cv-cvad-wlga-ca-2606
- DewiBrynJones/evals-speech-recognition-cy-en-2605
- DewiBrynJones/banc-trawsgrifiadau-bangor-2605
- DewiBrynJones/evals-speech-recognition-cy-en-2602
- DewiBrynJones/preprocessed-whisper-btb-cv-cvad-wlga-ca-2603
- DewiBrynJones/preprocessed-whisper-btb-cv-cvad-wlga-ca-2602
- DewiBrynJones/banc-trawsgrifiadau-bangor-2602
- DewiBrynJones/preprocessed-whisper-btb-cv-cvad-cven-wlga-ca-ec-2601
- DewiBrynJones/evals-speech-recognition-cy-en-2601
- DewiBrynJones/evals-speech-recognition-cy-en-2512
- DewiBrynJones/evals-speech-recognition-cy-en-2511
Techiaith demos on Hugging Face Spaces (4)
- microsoft_speecht5_finetuned_bu_tts_cy_enText-to-speech demo
- mimicMIMiC
- llama-3.2-1b-lora-cyChat with a Welsh Llama model
- READMEAccount information
DewiBrynJones demos on Hugging Face Spaces (3)
- Adnabod_Lleferydd_CymraegWelsh speech recognition with Wav2Vec2; build error at the time of the snapshot
- openWakeWordWake-word recognition; runtime error at the time of the snapshot
- trackioNo application file was available at the time of the snapshot
Machine translation
These are the dedicated translation models and translation memories appearing in the two accounts. The translation direction forms part of many model names: en-cy or eng-to-cym for English to Welsh, and cy-en or cym-to-eng for Welsh to English.
Techiaith translation models (5)
- techiaith/nllb-200-1.3B-ft-eng-to-cym-legislationEnglish to Welsh · legislation
- techiaith/nllb-200-1.3B-ft-cym-to-engWelsh to English
- techiaith/m2m100_1.2B-ft-cy-to-enWelsh to English
- techiaith/mt-dspec-health-en-cyEnglish to Welsh · health
- techiaith/mt-dspec-legislation-en-cyEnglish to Welsh · legislation
DewiBrynJones translation models (3)
- DewiBrynJones/bergamot-mt-cy-enWelsh to English
- DewiBrynJones/m2m100_1.2B-ft-cy-to-enWelsh to English
- DewiBrynJones/nllb-200-1.3B-ft-eng-to-cymEnglish to Welsh
Techiaith translation datasets (9)
- techiaith/bergamot-cy-en-translations
- techiaith/cardiff-university-tm-en-cy
- techiaith/llyw-cymru-en-cy-ogl
- techiaith/banc-trawsgrifiadau-bangor_mt
- techiaith/YouTube-Subtitles
- techiaith/cofnodycynulliad_en-cy
- techiaith/bydtermcymru-tm-en-cy
- techiaith/legislation-gov-uk_en-cy
- techiaith/cwm-taf-morgannwg-university-health-board-tm-en-cy
DewiBrynJones translation dataset (1)
Read the overview of Welsh translation tools →
Language models, language processing and evaluation
This section includes Welsh language models, a model that restores capitalisation, reasoning and safety evaluation datasets, and data for recognising Macsen users’ intentions.
Techiaith models — language models and language processing (4)
Techiaith datasets — language processing and evaluation (7)
- techiaith/toxigen-dataEvaluating toxic content
- techiaith/finepdfs-cy-errorsErrors in Welsh PDF data
- techiaith/llyw-cymru-cy-ogl
- techiaith/macsen_intent_parsingRecognising Macsen users’ intentions
- techiaith/wnli-cyTextual inference
- techiaith/COPA-cyCause-and-effect reasoning
- techiaith/mgsm_cyMathematical word problems
DewiBrynJones dataset — language processing (1)
- DewiBrynJones/macsen_intent_parsingRecognising Macsen users’ intentions
Techiaith collections
Hugging Face collections bring together related models and datasets. The DewiBrynJones account had no public collection on the snapshot date.
Public collections (7)
- Speech Recognition ModelsWelsh and bilingual speech recognition models
- Speech Recognition DatasetsSpeech recognition datasets
- Machine Translation ModelsMachine translation models
- Machine Translation DatasetsMachine translation datasets
- Evaluation DatasetsEvaluation datasets
- Large Language ModelsWelsh large language models
- Agile Cymru (Cymru–Breizh) ModelsWales–Brittany Linguistic AI Network models
How to use the catalogue
- Check each model or dataset card for its licence, provenance, version and intended use.
- Inclusion in this list does not mean AIsteddfod has verified the item’s quality. Our test suites page sets out what can currently be measured.
- When comparing models, record the exact repository name and date: several series contain versions from different periods or formats for different systems.
The original accounts: Techiaith models · Techiaith datasets · Techiaith Spaces · Techiaith collections · DewiBrynJones models · DewiBrynJones datasets · DewiBrynJones Spaces.
