Amazon’s new multimodal AI can produce cities out of pasta

Amazon has launched a new family of “understanding” multimodal AI models that can generate text, images and videos from different inputs.

Amazon, known collectively as Nova, says it is capable of “analyzing complex documents and videos, understanding charts and graphs, generating engaging video content, and developing sophisticated AI agents.”

Part of Nova is an AI image generator and a video generator. Amazon Nova Canvas is a “modern imaging model” that can produce “studio-quality” images and offers precise control over style and content. It has features like inpainting, outpainting and background removal.

Amazon Nova Reel can produce short videos from text input or image input. Amazon says it can “generate professional-quality video content for marketing, advertising and entertainment.” The company released an ad for a fake Nova Reel noodle brand, depicting a city made of noodles.

A number of other short clips were also shared (see below). The company emphasizes that “all Amazon Nova models have built-in security controls and creative content creation models have watermarking capabilities to promote the responsible use of AI.”

In its press release, Amazon says three understanding models are about to launch and a fourth will follow soon.

The first model is Amazon Nova Micro, a text-only model that works similarly to ChatGPT. The second is Amazon Nova Lite, a “low-cost” multimodal model that can process image, video and text input to generate text output. Finally, there’s Amazona Nova Pro, a “high-performance multimodal model.”

The fourth model, Amazon Nova Premier, will be its “most powerful multimodal model for complex reasoning tasks.” According to Amazon, Nova Premier should be available in early 2025.

The new Nova AI Foundation models will be available exclusively as part of the Amazon Bedrock model library on Amazon Web Services (AWS).

Nova AI was announced during the company’s AWS re:Invent conference in Las Vegas. The edge notes that the company also said it plans to build a massive AI computing cluster based on its Trainium 2 chips through its partnership with Anthropic.

“When completed, it is expected to be the world’s largest AI computing cluster available to Anthropic to date on which to build and deploy its future models,” Amazon added.