We wrote a new paper entitled “CoNeTTE: An efficient Audio Captioning system leveraging multiple datasets with Task Embedding”, available on arxiv!