In response to Google’s Gemini launch, Meta is introducing a standalone generative AI experience called Imagine with Meta on the web. This feature enables users to effortlessly create images simply by describing them in natural language.
Imagine With Meta, a Generative AI powered by Meta’s Emu image generation model, allows U.S. users to create high-resolution images from text prompts.
Similar to OpenAI’s DALL-E, Midjourney, and Stable Diffusion, this platform generates up to four images per prompt and is currently available for free (at least for now).
“We have been delighted to receive feedback from individuals on their use of imagine, Meta AI’s text-to-image generation feature, to produce entertaining and imaginative content during conversations”.
“Today, we are pleased to announce that access to imagine is being extended beyond chats,” Meta states in a blog post released this morning. “
Although our messaging platform is primarily designed for lighthearted and interactive exchanges, you now have the option to generate free images on the web as well.”
Given Meta’s previous issues with racially biased AI sticker generator, it raises concerns about the safeguards in place in Imagine with Meta to prevent similar incidents.
Although we didn’t have the opportunity to test the tool before its launch, we will closely monitor its performance as it reaches a wider user base.
At the initial launch, ImagineWithMeta will not feature live watermarks. However, Meta has committed to incorporating watermarks in the generated content shortly to enhance transparency and traceability.
Currently, there is already a visible watermark in place. These watermarks, although invisible, will be produced using an AI model and can be identified using a corresponding model developed by Meta. It is unclear whether the detection model will be made publicly available in the future.
Meta stated in their post that the watermarks are designed to withstand various image manipulations, including cropping, resizing, color changes (such as brightness and contrast adjustments), screen captures, image compression, noise, sticker overlays, and more.
They expressed their intention to implement invisible watermarking in many of their products using AI-generated images in the future.