EchoSight: Unified Embedding Space for Audio and Video

EchoSight explores a shared embedding space for audio and video, inspired by CLIP-style contrastive learning. The goal is to represent paired sound and visual content for cross-modal search and related retrieval tasks.