Description
Mainstream speech recognition covers a few dozen major languages, leaving minority languages untranscribable. Omnilingual ASR is Meta's open-source multilingual speech recognition for 1600+ languages.
It offers models from small to large and extends to new languages with a few examples.
1600+ languages:Long tail covered.
Sizes:Pick one.
Few-shot:New languages.
Open:Downloadable models.
It offers models from small to large and extends to new languages with a few examples.
Features
1600+ languages:Long tail covered.
Sizes:Pick one.
Few-shot:New languages.
Open:Downloadable models.
Screenshots
Tags:speech-recognition