Photographer: David Paul Morris/Bloomberg via Getty Images
© 2024 Blaze Media LLC. All rights reserved.
Meta releases its own AI that detects every object in an image, creates videos from text
April 07, 2023
Mark Zuckerberg's Meta has released its own artificial intelligence model capable of identifying objects within an image, along with other features such as generating complex images and videos from text, according to a report by Reuters.
A Meta blog post reportedly revealed the company's new AI "Segment Anything Model," called "SAM" for short, that can identify objects in images and videos. A demonstration showed typing the word "cat" to prompt SAM to draw boxes around the cats in a photo.
The release of the program comes amid AI platforms popping up everywhere, with Microsoft-backed ChatGPT facing backlash in Italy over a privacy breach that revealed users' private information, causing a temporary suspension of the service.
Popular cartoon "South Park" even wrote part of a recent episode using ChatGPT.
\u201cMeta unveils new AI model as CEO Mark Zuckerberg pushes to incorporate generative artificial intelligence into Meta's apps in 2023 https://t.co/P9lEWCberd\u201d— Reuters (@Reuters) 1680835800
Meta's patience in releasing its own AI may have paid off, however, as the company is set to release more advanced text-to-image and even text-to-video applications.
With "Make-A-Scene," users can generate images with text prompts, described as "Scene-Based Text-to-Image Generation with Human Priors."
This means that not only can the program create images with just text, it can also upgrade or enhance existing images. In a published video, a story is created using simple sketches and text prompts, identifying just a boat, sea, sky, and mountain.
Meta compared its programs with competitors in a research paper, showing the AI's capabilities to create cartoon-like sketches, realistic images, and even imaginary concepts like "a car with triangular wheels."
With the "Make-A-Video" studio, users can create moving video from text and a single or multiple images, or even create variations of existing video.
The very first sample is a video created from the text "A dog wearing a Superhero outfit with red cape flying through the sky."
The platform showcases "surreal" video types such as a "a teddy bear painting a portrait"; a "young couple walking in a heavy rain" in its "realistic" category; and a "stylized" video of "a table by a window with sunlight streaming through illuminating a pile of books."
The SAM model and data set will be available to download under a non-commercial license allowing the AI platform to be used for research purposes only.
Like Blaze News? Bypass the censors, sign up for our newsletters, and get stories like this direct to your inbox. Sign up here!
Want to leave a tip?
We answer to you. Help keep our content free of advertisers and big tech censorship by leaving a tip today.
Want to join the conversation?
Already a subscriber?
<p>Andrew Chapados is a writer focusing on sports, culture, entertainment, gaming, and U.S. politics. The podcaster and former radio-broadcaster also served in the Canadian Armed Forces, which he confirms actually does exist.</p>
andrewsaystv
Andrew Chapados
Andrew Chapados is a writer focusing on sports, culture, entertainment, gaming, and U.S. politics. The podcaster and former radio-broadcaster also served in the Canadian Armed Forces, which he confirms actually does exist.
more stories
Sign up for the Blaze newsletter
By signing up, you agree to our Privacy Policy and Terms of Use, and agree to receive content that may sometimes include advertisements. You may opt out at any time.
© 2024 Blaze Media LLC. All rights reserved.
Get the stories that matter most delivered directly to your inbox.
By signing up, you agree to our Privacy Policy and Terms of Use, and agree to receive content that may sometimes include advertisements. You may opt out at any time.