How up to date is this with the original whisper?

[ururk]

I'm noticing some major differences between the original whisper and whisper.cpp, which prevent me from using this project. Things like:

all audio transcribed as . or >> (no words)
getting special caption notations: [ BREAK ]

Those are some of the differences. I tried to make sure I was running with the same params, but I'm still not getting similar output. Any suggestions?

How close does whisper.cpp match openai/whisper? Things that "concern" me are fixes in the original whisper project that address infinite loops.