While working on audio chunking I noticed that sometimes language detection is off. It can happen that language is detected as <|nocaptions|> which might result in a whole 30s segment being discarded.
This PR fixes that by defaulting to "en" when the language detection is confused.
While working on audio chunking I noticed that sometimes language detection is off. It can happen that language is detected as
<|nocaptions|>
which might result in a whole 30s segment being discarded.This PR fixes that by defaulting to "en" when the language detection is confused.