Spaces:

Duplicated from  whisper-event/whisper-demo

softcatala

/

whisper-demo-catalan

like 1

Runtime error

Hi @jordimas - Thank you for applying for a community grant. Can you please provide us with your models WER and how it compares with whisper medium/ large-v2's zero shot WER?

All for Catalan language. Here you are:

• Small: https://huggingface.co/softcatala/whisper-small-ca - WER de 8.56
• Meidum: https://huggingface.co/softcatala/whisper-medium-ca - WER de 8.32

Evaluation on CV11 on test dataset.

For large-V2 on VC9 dataset, OpenAI reports a WER of 14.1