results from 101 sample tv news broadcasts spanning 50 countries in both transcription and translation in tiny, small, medium and large models

[kalevleetaru]

For those interested in how Whisper performs on an archive of content from across the world, we've applied it to 101 sample television news broadcasts spanning 50 countries in both transcription and translation tasks across the Tiny, Small, Medium and Large models. You can view all of the outputs to see how it performed and for a number of the broadcasts you can compare against the original show to see how well it performed and where it struggles.

https://blog.gdeltproject.org/applying-openais-whisper-asr-to-101-sample-tv-news-broadcasts-spanning-50-countries/