Google's new AI tool turns text into HD video

It was only a matter of time before Google entered the field of AI-assisted video creation, and the start is with the research version of IMAGEN Video, which is reminiscent of Meta's Make-A-Video system.

IMAGEN Video is fed with short written descriptions given by the user and then the AI ​​algorithm converts it into an image and finally into a short HD video. The results are truly impressive, although for now the tool is in embryonic development and Google has several reservations about opening it up to the world, fearing potential malicious use.

The new AI tool proceeds to create the HD video by analyzing the written description in 8 stages. It starts with a low resolution demo (24 x 48 / 3fps) which then scales exponentially (by a factor of 18x) to reach a final resolution of 1280 x 768 / 24fps lasting 5.3s. There are plenty of stylistic models (eg video production of works by famous painters) and the algorithm is constantly trained with new material to provide even more accurate results.

In addition to IMAGEN VIDEO, the company is also launching a text-to-video tool, called Phenaki, which has the ability to create longer videos using more detailed commands. Also, the DreamFusion tool creates 3D models from text.