Pasta City
ABC (USA)

General manager AmazonAndy Jassy took the stage at the company’s re:Invent conference on Tuesday to demonstrate six new models for generating text, images and video that require Amazon Nova.

This new family of multimodal generative AI includes Nova Micro, a text-based model designed for low-cost, low-latency responses; Nova Lite is a low-cost multimodal model for processing image, video and text inputs; and Nova Pro, a general-purpose multimodal model that combines “precision, speed and cost for a wide range of applications,” according to the company. Nova Premier is “…multimodal models for complex logic problems,” Nova Canvas is a dedicated text-to-image engine, and Nova Reel is designed specifically for video creation.

Recommended Videos

Text models are optimized for 15 different languages. Micro-offers context window for 128,000 tokens, while Lite and Pro can handle up to 300,000 tokens (about 225,000 words or 30 minutes of video). The company plans to expand the context windows of its largest models to 2 million tokens early next year.

Canvas allows users to create and edit images using natural language prompts. The Reels, which will compete with Gen-3 Alpha, Kling and Dall-E 3can create clips up to six seconds long from both text prompts and reference images. The video generator also offers camera motion control, including panning and zooming.

Pasta City created with Amazon Nova Reel from Amazon Ads

“We have continued to work on our own advanced models,” Jassy told the assembled crowd, “and these advanced models have made tremendous progress in the last four or five months. And we realized that if we find value in them, you will probably find value in them.”

Jassy also says the models are among the cheapest to run and fastest in their class, although the company has yet to release benchmark data to back up those claims. “We’ve optimized these models to work with proprietary systems and APIs, so you can perform multiple automated, orchestrated steps (agent behavior) much more easily using these models,” he said. “So I think they’re very compelling.”

Micro, Lite, and Pro models (as well as Canvas and Reels) are currently available to AWS customers. The premiere is scheduled for the first quarter of 2025.

Source: Digital Trends

Previous articlePersonal artificial intelligence assistants will help astronauts on long flightsScience and technology04 December 2024 22:03
Next articleThe director of OpenAI said he is not afraid of Musk’s political influence
I am Garth Carter and I work at Gadget Onus. I have specialized in writing for the Hot News section, focusing on topics that are trending and highly relevant to readers. My passion is to present news stories accurately, in an engaging manner that captures the attention of my audience.

LEAVE A REPLY

Please enter your comment!
Please enter your name here