Not to be outdone by the launch of Google’s Gemini, Meta is launching a new standalone generative AI experience on the web, Imagine with Meta, which allows users to create images by describing them in natural language.
Similar to OpenAI’s DALL-E, Midjourney, and Stable Diffusion, Imagine with Meta, which works with Meta’s existing Emu imaging model, creates high-resolution images from text prompts. It’s free to use (at least for now) for US users and generates four images per message.
“We’ve enjoyed hearing from people about how they use Imagine, Meta AI’s text-to-image generation feature, to create fun and creative content in chats. Today, we are expanding access to imagination outside of chats,” writes Meta in a blog post published this morning. “While our messaging experience is designed for more fun, back-and-forth interactions, now you can also create free images on the web.”
Now, Meta’s image generation tools have landed the company in trouble in the recent past (see: Meta’s racially biased AI sticker generator), which makes this writer wonder if there are safeguards in Imagine with Meta to prevent history from repeating itself. We didn’t have a chance to test the tool before its release, but rest assured we’ll be keeping an eye on it as Imagine with Meta reaches more users.
They won’t be available at first, but Meta has promised to start adding watermarks to content generated by Imagine with Meta in the coming weeks for “greater transparency and traceability.” (There is already a visible watermark). The watermarks, which are invisible, will be generated with an AI model and can be detected with the corresponding model, says Meta. It is not known if the detection model will ever be made public.
“[The watermarks are] resistant to common image manipulations such as cropping, resizing, color changing (brightness, contrast, etc.), screenshots, image compression, noise, sticker overlays, and more,” Meta stated in the post. “Our goal is to add invisible watermarks to many of our products with AI-generated images in the future.”
Watermarking techniques for generative art are not new. French startup Imatag offers a watermarking tool that it claims is not affected by resizing, cropping, editing or compressing images. Another company, Steg.AI, uses an AI model to apply watermarks that survive resizing and other edits. Microsoft and Google have adopted AI-based watermarking standards and technologies, while elsewhere, Shutterstock and Midjourney have agreed on guidelines for embedding markers that indicate their content was created using a generative AI tool.
But pressure is mounting on tech companies to make it clearer that works were generated by AI, particularly in light of the spate of Deepfakes from the Gaza war and AI-generated child abuse images that bypass filters.
Recently, the Cyberspace Administration of China issued regulations requiring generative AI vendors to mark up generated content (including text and image generators) without affecting user use. And in recent US Senate committee hearings, Senator Kyrsten Sinema (I-AZ) emphasized the need for transparency in generative AI, including the use of watermarks.