EmbeddingGemma 2 in FiftyOne: search video by text, sound, or webcam

Google DeepMind released EmbeddingGemma 2, an open 740M-parameter model that maps text, images, video, and audio into a single 768-dimensional space. A FiftyOne walkthrough builds three LanceDB indexes over 10,000 VGGSound clips to search video by description, soundtrack, or a live recording.

• 740M-parameter model under Apache 2.0 with an 8K-token context
• Matryoshka training allows truncating vectors to 512, 256, or 128 dimensions
• Three vector fields and LanceDB indexes over 10,000 VGGSound clips
• Search by text, uploaded media, microphone, or webcam