Google's open 740-million-parameter model turns text, images, audio, and video into matching number maps.
It gives words, photos, and sounds the same school ID card. They can all enter the same “looks alike” line.
It can search across media on your device. It also helps knowledge tools find related content.
Embedding
It turns text, images, audio, and more into one shared number map.
Gemma
It belongs to Google's open Gemma model family.
Multimodal AI
It uses one format for many types of content.
Vector search
Its vectors help find content with similar meaning.