My speech-to-text model produces accurate transcripts when tested in a quiet office.However, when I try to use it in public places, accuracy drops sharply.Background noise causes words to be skipped or misheard.The model feels fragile outside controlled ...