18 upvotes
EmbeddingGemma 2 unifies text, image, audio, video into one embedding
What is EmbeddingGemma 2: Unified and Efficient Multimodal Representations?
EmbeddingGemma 2 is an open-source model that creates a single vector for text, images, audio, video, and mixed-modality inputs. It achieves this with simplified per-layer embeddings and a multi-stage curriculum that keeps representations efficient.
Who it is for
Anyone building search, recommendation, or cross-modal retrieval pipelines can replace multiple specialty encoders with this single model.