Google has revolutionized the field of artificial intelligence with its new model, VideoPoet, capable of generating videos from texts.
This tool stands out for its high quality in terms of detail and continuity, surpassing the company’s previous models. VideoPoet can maintain a consistent layout across its elements over several seconds, reflecting a significant advancement in AI-based video generation technology.
In addition to its ability to create videos, VideoPoet can perform multiple simultaneous tasks, such as converting videos to audio, stylizing videos, and transforming images into videos. One of the most impressive aspects of VideoPoet is its consistency; can define each second of the video based on the previous one, resulting in a nearly identical representation of elements and characters over time. Google has demonstrated the potential of VideoPoet through creative examples, such as an origami fox in a forest, a medical squirrel, and a shark shooting laser beams. In addition, they have published a short film starring a raccoon that travels around the world and into space, thus demonstrating the narrative capacity of AI.
For this project, Google combined the capabilities of VideoPoet with Bard, another of its AI models. Bard was used to write the script for the short film, including a list of prompts for the video generator, which VideoPoet then transformed into real footage. This advance represents a great step in the generation of audiovisual content through artificial intelligence.