or Google's multimodalembedding@001 Model

A Vertex AI foundation model that generates unified 1408-dimension vectors from text, image, and video inputs for cross-modal search.

Google's multimodalembedding@001 model (part of the Vertex AI ecosystem) maps diverse data types into a shared vector space. It processes text (up to 32 tokens), images (standard formats), and video (up to 120 seconds) to enable high-performance applications like semantic image retrieval and video content recommendation. By outputting a consistent 1408-dimensional embedding, it allows developers to calculate cosine similarity across different media formats without separate specialized encoders.

https://cloud.google.com/vertex-ai/generative-ai/docs/embeddings/get-multimodal-embeddings