README.md

---
license: apache-2.0
base_model: openai/whisper-large-v3
tags:
- generated_from_trainer
- verbatim
metrics:
- wer
model-index:
- name: whisper-large-v3-ft-btb-cv-cy
  results: []
datasets:
- techiaith/banc-trawsgrifiadau-bangor
- techiaith/commonvoice_18_0_cy
- techiaith/commonvoice_vad_cy
- cymen-arfor/lleisiau-arfor
language:
- cy
- en
pipeline_tag: automatic-speech-recognition
---

# whisper-large-v3-ft-verbatim-cy-en

This model is a version of [openai/whisper-large-v3](https://huggingface.co/openai/whisper-large-v3) finedtuned with 
transcriptions of Welsh language spontaneous speech from 
[Banc Trawsgrifiadau Bangor (btb)](https://huggingface.co/datasets/techiaith/banc-trawsgrifiadau-bangor) and
[Lleisiau Arfor](https://huggingface.co/datasets/cymen-arfor/lleisiau-arfor) as well as recordings of read speech 
from [Welsh Common Voice version 18 (cv)](https://huggingface.co/datasets/techiaith/commonvoice_18_0_cy) and 
[Welsh Common Voice Vad Segments](https://huggingface.co/datasets/techiaith/commonvoice_vad_cy) for additional training. 

As such this model is suitable for more verbatim transcribing of spontaneous or unplanned speech. It achieves the 
following results on the [Banc Trawsgrifiadau Bangor'r test set](https://huggingface.co/datasets/techiaith/banc-trawsgrifiadau-bangor/viewer/default/test)

- WER: 28.99
- CER: 10.27


## Usage

```python
from transformers import pipeline

transcriber = pipeline("automatic-speech-recognition", model="techiaith/whisper-large-v3-ft-btb-cv-cy")
result = transcriber(<path or url to soundfile>)
print (result)
```

`{'text': 'ymm, yn y pum mlynadd dwitha 'ma ti 'di... Ie. ...bod drw dipyn felly do?'}`