The text prompt ‘a small cactus wearing a straw hat and neon sunglasses in the Sahara desert’ was entered to create this image. — AFP Relaxnews
Meta has announced that it has developed its own artificial intelligence technology, capable of creating images of stunning quality from simple written prompts. In doing so, the American tech giant is entering a field that has gained considerable momentum in recent months, with the emergence of tools such as Dall-E and Midjourney, and the development of Adobe Firefly.
Having recently made headlines with the launch of its Twitter rival, Threads, Meta has now presented its latest project, CM3Leon. Behind this strange-looking name (pronounced "chameleon") lies a new image-generating artificial intelligence technology.
This multimodal model makes it possible to obtain images from text descriptions and vice versa. Despite a still limited learning base and far less computational power than the competition, Meta boasts the ability to produce more coherent imagery, particularly when it comes to detail, thus responding perfectly to the prompts it receives.
To illustrate its abilities, Meta has posted images generated from original prompts, such as "a small cactus wearing a straw hat and neon sunglasses in the Sahara desert," or "a raccoon main character in an Anime preparing for an epic battle with a samurai sword. Battle stance. Fantasy, Illustration," or even "A stop sign in a Fantasy style with the text '1991.'"
In order to refine the quality of the image, corrections can be made by means of simple adjustments. For a portrait, for example, you could ask the tool "add a pair of sunglasses," "add face paint" or even "make them look like someone from 100 years ago."
Image-generation models are becoming increasingly sophisticated. Still, users will have to wait a little longer to try this one out, since Meta has yet to provide a timetable for its availability.
In recent months, Meta has been particularly active in the field of artificial intelligence, unveiling its artificial intelligence model LLaMA (currently reserved for academic researchers), its Voicebox studio, capable of transforming text into speech in different languages, and the MusicGen tool for generating music from simple text description. – AFP Relaxnews
