Canvas allows users to generate and edit images, while Reel creates videos up to six seconds long from prompts or reference images. AWS CEO Andy Jassy stated that the Nova models are among the fastest and least expensive to run. He also revealed that AWS is working on a speech-to-speech model for Q1 2025 and an "any-to-any" model for mid-2025. However, AWS remains vague about the data used to train its generative models.
Key takeaways:
- Amazon Web Services (AWS) announced a new family of multimodal generative AI models called Nova, which includes four text-generating models: Micro, Lite, Pro, and Premier.
- Nova also includes an image-generation model, Nova Canvas, and a video-generating model, Nova Reel, both of which launched on AWS.
- The Nova models are optimized for 15 languages and have varying capabilities, with the Premier model being the most capable and designed for complex workloads.
- AWS is also working on a speech-to-speech model for Q1 2025, and an “any-to-any” model for around mid-2025, which will be able to input and output text, speech, images, or video.