New York (EFE).- Meta (parent company of Facebook, Instagram and WhatsApp) launched this Friday CM3leon a generative artificial intelligence (AI) tool capable of generating an image from a text and vice versa.
In a statement, the company gives several examples: if one types “a small cactus with a straw hat and neon sunglasses in the Sahara desert”, an image with that description is generated.
Another function consists of editing an image from an indication with a text, such as “modify the color of the sky” in a photo or “put a mustache” on Johannes Vermeer’s painting of “Girl with a Pearl Earring”; In addition, it allows you to ask the AI to describe a photo with words.
Since AI became popular late last year, several companies have released such tools, and Meta notes that CM3leon is better than some of them.
“When comparing performance against the most widely used imaging benchmark (Zero-shot MS-COCO), CM3Leon achieves an FID (Fréchet Starting Distance) score of 4.88, establishing a new state of the art in imaging. generation of text to image and surpassing Google’s model of text to image'(Parti)”, the company notes in the statement.
However, Meta does not mention OpenAI’s Midjourney or Dall-e, the most popular tools for this kind of task so far.
