By Marina Avilés, on 18 March 2025
Nova is a family of artificial intelligence models launched by Amazon. With Nova, you can generate AI tools for text, image, and video, all while easily integrating them into the AWS environment where they are hosted. It’s crucial to understand that Nova isn’t a standalone AI tool but rather a model or a technology capable of powering a wide range of AI applications.
With Amazon Nova, the brand hopes to join the AI race–already dominated by players like OpenAI, Google, and Microsoft. Amazon currently controls over 30% of the global cloud market, leading in server and cloud service sales. It now seeks to leverage AI to maintain its competitive edge.
This article breaks down everything we know about Nova (so far) and how you can use it for your business and marketing efforts.
The 6 AI Models in the Nova Family
The Nova family consists of six AI models:
Text-generation models: Micro, Lite, Pro, and Premier.
Creative content models: Nova Canvas and Nova Reel.
Amazon’s goal is to gradually improve each of these models, but the company claims they are already highly capable, cost-effective, fast, and easy to use.
Amazon Nova Micro
This model generates text from text inputs. It’s the most cost-effective and fastest model in the family, ideal for real-time analytics tools and interactive chatbots. For reference, Nova Micro can write approximately 25 pages of text in under a minute.
Amazon Nova Lite
A multimodal model, Nova Lite accepts text, images, and videos as inputs but generates text exclusively. It strikes an excellent balance between precision and speed, making it a great option for reasoning tasks and translations.
Amazon Nova Pro
Also multimodal, Nova Pro offers greater precision and speed compared to Lite. It’s the most advanced text-based model in the Nova family and is perfect for developing tools like video summarization, workflow automation, and software development aids.
Amazon Nova Premier
This is the most sophisticated multimodal text-generation model, capable of handling complex tasks and creating custom models. As of this writing, Nova Premier is not yet available but is expected to launch in the first quarter of 2025. Depending on when you’re reading this, it may already be live.
Nova Canvas
This AI model specializes in editing and generating images from text and existing images. It’s particularly useful for removing unwanted elements from backgrounds, altering colors, and filling in missing parts of an image. Nova Canvas is an excellent tool for creating marketing visuals.
Nova Reel
Nova Reel can generate short videos up to six seconds long based on text prompts or reference images. Amazon’s ultimate objective is for Nova Reel to create videos lasting up to two minutes.
How to Access Nova
Nova is available via AWS Bedrock, a managed service by Amazon Web Services that simplifies the creation and optimization of generative AI applications. Businesses already using AWS can easily integrate Nova models directly into the AWS platform.
According to Rohit Prasad, Senior Vice President of AI at Amazon:
"Our new Nova models are designed to help internal and external developers tackle challenges through compelling, cost-effective AI with high customization potential and low latency."
Amazon claims that all Nova text-generation models are 75% more affordable and faster than the top-paid models in their respective categories currently on the market.
Retailers and advertisers, in particular, can benefit from Nova as a tool to boost creativity in product catalogs and paid campaigns.
A Customized AI Model
Nova models are highly personalized, as they are designed to learn from the data provided by the users. This enables Nova to deliver accurate results while understanding customer priorities across text, image, or video formats.
The Future of the Nova Family
Amazon plans to expand the Nova family in 2025, introducing two new models:
A voice-to-voice AI model.
A fully multimodal model capable of processing inputs in any format (text, image, video, or audio) and generating outputs in any of these formats—a true multimodal-to-multimodal system.
