README.md

metadata

library_name: transformers
license: apache-2.0
base_model: openai/whisper-large-v3
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: whisper-large-v3-ft-cv-cy-en
    results: []
datasets:
  - techiaith/commonvoice_18_0_cy_en
language:
  - cy
  - en
pipeline_tag: automatic-speech-recognition

whisper-large-v3-ft-cv-cy-en

This model is a fine-tuned version of openai/whisper-large-v3 on the techiaith/commonvoice_18_0_cy_en dataset. Both the English and Welsh data have been used to fine-tune the whisper model for transcribing both languages as well as improved language detection.

It achieves a success rate of 98.86% for language detection on recordings from a Common Voice bilingual test set

While, it achieves the following WER results for transcribing using the same test set:

• Welsh: 26.20
• English: 15.37
• Average: 20.70

N.B. the desired transcript language is not given to the fine-tuned model during testing.

Usage

from transformers import pipeline

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

{'text': 'Mae hen wlad fy nhadau yn annwyl i mi.'}