Not to be outdone by Google, Meta launches AI that creates images using natural language

Similar to OpenAI’s DALL-E, Midjourney, and Stable Diffusion, the new application, called “Imagine with Meta,” is powered by the social media giant’s existing Emu image generation model.

Meta is releasing the algorithm for free to users in the US, which is capable of generating four different high-resolution images for each text prompt.

However, with this AI image generator, Meta will continue to face recent problems, such as the company's AI sticker generator being racially biased.

Parent company Facebook also pledged to begin watermarking content created by Imagine with Meta in the coming weeks to “increase transparency and traceability.”

These watermarks will be generated using an AI model and will be detectable by a matching model, Meta said, without revealing further information about the matching recognition model.

“The watermark is resilient to common image editing operations such as cropping, resizing, color changes (brightness, contrast, etc.), screenshots, image compression, noise, sticker overlays, and more,” Meta said in the post. “We look forward to bringing invisible watermarks to more of our products with AI-generated images in the future.”

Watermarking for creative art is not new. Several startups such as Imatag (France) and Steg.AI have watermarking solution tools that they claim are unaffected by photo editing including resizing, cropping, editing or compression.

Microsoft and Google have also adopted AI-based watermarking standards and technology, while elsewhere, Shutterstock and Midjourney have agreed to principles that embed markers indicating their content was generated by a synthetic AI tool.

Recently, the Cyberspace Administration of China issued regulations requiring synthetic AI providers to watermark the content they generate — including text and image generators — without affecting user use.