song-describer-dataset

The Song Describer Dataset (SDD) serves as a corpus of approximately 1.1k audio captions for 706 music recordings, specifically designed for evaluating music-and-language models in tasks such as music captioning and text-to-music generation. The dataset includes detailed information on each recording and showcases various notable features like permissive licensing, diverse caption lengths, and extensive vocabulary, supporting researchers in advancing multimodal audio analysis.