SpeechCaps: Advancing Instruction-Based Universal Speech Models with Multi-Talker Speaking Style Captioning

Published in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2025