How close do you aim for when it comes to drop-in vs whisper.cpp? Are timestamps per word and character something aimed for? How about multi-lingual transcription or hallucination suppression?
The github page doesn't seem to go into depth on these orthogonal topics. May have missed it.