Amazon Web Services has introduced Nova, a new family of artificial intelligence models designed for text, image and video generation. The announcement was made by CEO Andy Jassy at the AWS re:Invent conference.
The Nova family consists of four text-generating models: Micro, Lite, Pro, and Premier. Micro, Lite, and Pro are immediately available to AWS customers, while Premier is scheduled for release in early 2025. The models vary in capabilities, with Micro being text-only and the others able to process multiple types of input including images and video.
The text models feature large context windows, with Micro handling 128,000 tokens and Lite/Pro supporting 300,000 tokens. According to Amazon, these models are optimized for 15 languages, with a primary focus on English. The company states these models are among the fastest and most cost-effective in their category.
Two additional specialized models were announced: Nova Canvas for image generation and Nova Reel for video creation. Canvas allows users to generate and edit images through text prompts, while Reel can create six-second videos with features like camera panning and zooming.
Amazon emphasizes built-in safety measures across the Nova family, including watermarking and content moderation. The models are available through AWS Bedrock, Amazon’s AI development platform, where they can be fine-tuned for specific uses.
In terms of pricing, Nova Micro is positioned as one of the market’s least expensive options at 3.5 cents per million input tokens. The company has stated that future developments will include a speech-to-speech model and an “any-to-any” modality model, both planned for 2025.
Sources: TechCrunch, VentureBeat, Simon Willison