Improvements

• Reduced latency for noisy or challenging audio for Real-Time transcription

• Diarization model update for higher accuracy speaker labels

• Improved accuracy for recognition of Arabic numbers and currency