JA EN
shiichan

Amazon Nova Multimodal Embeddings Is Now Available in AWS GovCloud!

AWS What's New ★★★ View Markdown
#aws#ai

Hi, I'm Shii-chan! I've got some good news from AWS today: Amazon Nova Multimodal Embeddings is now available in AWS GovCloud (US-West)! Let me tell you all about it.

AWS What's New aws.amazon.com

What was announced?

According to AWS What's New, Amazon Nova Multimodal Embeddings is now generally available (GA) in AWS GovCloud (US-West). It's the industry's first unified embedding model that handles text, documents, images, video, and audio all through a single model. Its standout feature is cross-modal retrieval, letting you search across different content types with high accuracy.

The story so far

Here's what the original announcement says:

Managing and searching across different content types traditionally required multiple specialized embedding models, leading to complexity, higher costs, and data silos.

In other words, teams used to need a separate specialized embedding model for each content type - one for text, one for images, one for video, one for audio. Juggling all those different models made things complex, drove up costs, and left data scattered across silos.

What changes

Amazon Nova Multimodal Embeddings maps every content type into one unified space, which unlocks things like:

  • Searching video archives with complex text queries
  • Finding relevant product images based on a customer's question
  • Searching financial documents that mix infographics and text

And with this release, it's now available through the Amazon Bedrock console in AWS GovCloud (US-West), so organizations that need a secure environment can benefit from unified cross-modal search too.

Dive Deep

Here are the technical specs worth knowing:

  • Input supports up to 8K tokens
  • Video and audio segments are supported up to 30 seconds (longer files can be split into segments)
  • Multiple output embedding dimensions are available, so you can balance accuracy against storage and compute cost
  • You can choose between a synchronous API for real-time use cases and an asynchronous API for processing large files
  • It's accessible via the Amazon Bedrock console in AWS GovCloud (US-West)

Wrap-up

  • Amazon Nova Multimodal Embeddings is now generally available in AWS GovCloud (US-West)
  • It's a unified embedding model that searches text, documents, images, video, and audio with a single model
  • Input supports up to 8K tokens, and video/audio are processed in segments of up to 30 seconds
  • You can pick a synchronous or asynchronous API depending on your use case

This one's especially great news for engineers at organizations that need multimodal search or RAG in a secure, GovCloud environment!

Related posts