An open-source audio understanding model supporting speech recognition, environmental sound analysis, music understanding, time-aware QA, and complex