AI development has since progressed from working with text to creating increasingly complex image outputs, and now videos. At the Google I/O event, the annual developer conference, Google introduced AI models for diverse tasks. Among them is Veo for video editing. It generates super realistic content over a minute's duration, and is designed to help creatives bring their ideas to life.
Veo is a next-level tool for professional filmmaking
Professional video production requires equipment, actors, and editing skills — all of which cost money. Google's Veo can potentially allow anyone to create any video with just a text prompt. The tech giant announced the AI model at the previously announced Google I/O conference on May 14. They claim that it can create 1080p resolution videos, which is basically high-definition quality.
So, you could describe a bustling city like New York and Veo would make a video that captures people rushing to and fro. You could also request a panoramic view of the Grand Canyon at sunset — or alpacas in sweaters dancing to a beat, like Google has demonstrated in Veo’s preview video.
The AI isn't entirely new. It builds on Google's previous work in video generation, including Imagen-Video, VideoPoet, and Lumiere.
Google is collaborating with top filmmakers in the movie industry
In understanding how filmmakers and creators actually go through the creative process, Google has invited a range of creators to experiment with the Veo model. We see Donald Glover and his Gilga team testing the feature in a video Google has shared on YouTube and the announcement blog post.
For now, Veo is not yet available to the public. However, Google is asking creators to join a waitlist and try the AI video editor through their VideoFX program. It's another new and experimental feature that's still locked within Google Labs. In the future, they'll expand access to Veo's capabilities, and may integrate some of its features into YouTube Shorts and other existing apps or services.
Google's new AI models will not generate human monstrosities
In other news, Google is also working on the Imagen 3 AI model. Most models struggle with creating perfectly realistic human hands in their early stages. You can't really blame them when you consider that they're not human themselves, and hands are among features with delicate details. You have to consider the number of fingers, joints, and wrinkles.
Imagen 3 is Google's most advanced system for generating images based on textual prompts. It produces images with a very high level of detail, making them appear photorealistic and lifelike. You'll see less blurry areas, inconsistencies, or nonsensical elements. You can barely even tell that the sample images Google shared on its blog post are not real.
The Music AI Sandbox is also in the works, which was developed with YouTube and artists like Björn from ABBA and Wyclef Jean. Like Veo, these tools are still under testing and creators need to join a waitlist to see how they works. Personally, I can't wait to use Veo and bring my personal collection of poems to life.
