Meta has launched a new web-based generative AI experience called Imagine with Meta, in order to compete with Google's Gemini. It allows users to generate images using natural language descriptions.

Similar to OpenAI's DALL-E, Midjourney, and Stable Diffusion, Imagine with Meta utilizes Meta's existing Emu image generation model to create high-resolution images from text prompts. It is currently available for free to users in the United States, with four images generated per prompt.

"We are excited to see people using Imagine, Meta AI's text-to-image feature, to create interesting and creative content in chats. Today, we are expanding Imagine's accessibility beyond chats," Meta said in a blog post this morning. "Our messaging experience is meant to be fun and interactive, but now you can also generate images online for free."

Although they are not immediately available, Meta promises to add watermarks to the content generated by Imagine with Meta in the coming weeks to achieve "greater transparency and traceability." The watermark, which is invisible and generated by an AI model, can be detected by another model, according to Meta. It is currently unknown when the detection model will be made public.

"Watermarks can withstand common image manipulations such as cropping, scaling, color changes (brightness, contrast, etc.), screenshots, image compression, noise, sticker overlays, etc.," Meta stated in the article. "Our goal is to use invisible watermarks in many of our future products, all of which are generated using AI."

Watermarking technology for generative art is not new. French startup Imatag offers a watermarking tool that claims to be unaffected by image scaling, cropping, editing, or compression. Another company, Steg.AI, uses an AI model to add watermarks that can withstand scaling and other edits. Microsoft and Google have both employed AI-based watermarking standards and technologies, while elsewhere, Shutterstock and Midjourney have agreed to embedded tagging guidelines, indicating that their content is generated using generative AI tools.