The AI meta tool cannot be drawn without user assistance
Meta, owned by Facebook, showed off a new artificial intelligence tool for generating its own images, which is not yet available to the public.
AI-generated images are now being used to create silly memes by Craiyon, and now Meta is getting into the ring with its own image generator, but more focused on creating pieces of art.
The new tool is called AI Make-A-Scene, and it opens in a new tab and uses a combination of text prompts and schematic diagrams to create an image of 2048 x 2048 pixels. Meta demonstrated the tool in a video in which the company asks several artists to create pieces of art.
Meta added graphics as part of the image-making process because textual prompts alone don’t do a great job of reflecting the artist’s intent, he says. The company provides an example of a tool request to create an image of a zebra cycling. With just a text vector, the zebra in the image might be too big or face the wrong direction, while the diagrams give a structure to follow.
The images created by the tool are more faithful to the outline than the text vector, which means that they are not drawn without the help of the user. According to developer tests, 99.54% of generated images were found to “align perfectly” with the outline. The tool accounted for 66.3% of text claims.
It’s not known when the tool will be available to the public, and the meta tool is meant to be used to help users with their own creative efforts or marketing campaigns, Tekradar said.
The new tool is called AI Make-A-Scene, and it opens in a new tab and uses a combination of text prompts and schematic diagrams to create an image of 2048 x 2048 pixels. Meta demonstrated the tool in a video in which the company asks several artists to create pieces of art.
Meta added graphics as part of the image-making process because textual prompts alone don’t do a great job of reflecting the artist’s intent, he says. The company provides an example of a tool request to create an image of a zebra cycling. With just a text vector, the zebra in the image might be too big or face the wrong direction, while the diagrams give a structure to follow.
The images created by the tool are more faithful to the outline than the text vector, which means that they are not drawn without the help of the user. According to developer tests, 99.54% of generated images were found to “align perfectly” with the outline. The tool accounted for 66.3% of text claims.
It’s not known when the tool will be available to the public, and the meta tool is meant to be used to help users with their own creative efforts or marketing campaigns, Tekradar said.