19-7-2023 (SAN FRANCISCO) Meta, the American tech giant behind Facebook and Instagram, has developed its own artificial intelligence technology called CM3Leon that generates high-quality images from simple written prompts. This move has put the company in the rapidly growing field of AI-assisted image generation, which has seen a surge in interest lately, with the emergence of tools like Dall-E and Midjourney, and the development of Adobe Firefly.
CM3Leon, pronounced “chameleon,” is a multimodal model that can generate images from text descriptions and vice versa. Despite still having a limited learning base and less computational power than its competitors, Meta claims that the technology can produce more coherent and detailed images that perfectly match the given prompts.
To demonstrate its capabilities, Meta has shared images generated from original prompts, such as “a small cactus wearing a straw hat and neon sunglasses in the Sahara desert,” “a raccoon main character in an anime preparing for an epic battle with a samurai sword. Battle stance. Fantasy, Illustration,” and “a stop sign in a fantasy style with the text ‘1991.’”
Users can refine the quality of the generated images by making simple adjustments. For instance, for a portrait, users can ask the tool to “add a pair of sunglasses,” “add face paint,” or “make them look like someone from 100 years ago.”
Meta has been particularly active in the field of artificial intelligence lately, unveiling various AI models such as LLaMA, Voicebox studio, and MusicGen. LLaMA is currently only available for academic researchers, while Voicebox studio allows users to transform text into speech in different languages. MusicGen, on the other hand, generates music from simple text descriptions.