Amazon has introduced a new AI designed to assist business customers. The online giant has been active in the AI sector for some time. After releasing the shopping assistant Rufus for Amazon customers at the end of October, Amazon is now focusing on vendors who sell products or advertise on its platform.
Amazon has announced six versions of a new AI model called Nova. These versions differ in functionality and complexity. For instance, Amazon Nova Micro is a text-based model optimized for speed and low cost. Amazon Nova Lite is also cost-efficient but can process text, images, and videos. There are also Pro and Premier versions, which are multimodal models capable of solving complex tasks.
In addition, Amazon Nova Canvas is designed for image generation, and Amazon Nova Reel is for video generation. Most models are now available in Amazon Bedrock in the USA, except for Amazon Nova Premier, which will be available in the first quarter of 2025. It’s unclear when these models will be available elsewhere. All Nova AI models understand 200 languages. Amazon Nova Lite, Micro, and Pro are expected to be about 75% cheaper than comparable AI models available through Amazon Bedrock.
Amazon Nova is primarily aimed at sellers, advertisers, and businesses. According to Amazon, these AI models are designed to “remove barriers for sellers and advertisers and enable a new level of creativity and innovation.” This includes video advertising, as demonstrated by a clip showcasing a fictional pasta brand with a city made entirely of pasta generated by the AI.
Another application is creating descriptions from videos. For example, Amazon Nova was shown a clip of a football game without sound, and the AI described what it “saw.” It could identify team colors and specific plays. Advertisers could use this feature to translate their videos into text and create suitable advertising or social media content.
Looking ahead, Amazon has additional plans for Nova in 2025. The AI model will have two more variants, which currently do not have final names. One will support language as input and output, while the other will be a multimodal model capable of generating multimodal output.
Amazon promises that the language model will even recognize non-verbal inputs, such as tone and speech rhythm. The multimodal model will serve as the foundation for AI agents. These agents, powered by Amazon Nova, will be able to perform multiple tasks in sequence or even operate devices independently.
Amazon continues to innovate in the AI field, aiming to provide advanced tools for businesses and advertisers to enhance their operations and creativity.