In September this year, Meta announced Meta AI, a virtual assistant powered by generative AI, for Instagram, Messenger, and WhatsApp. This chatbot features a tool called Imagine which can generate images from text prompts. The company made Meta AI and Imagine available to the public last month. Now, Meta has announced that it is expanding the availability of Imagine outside the chatbot and making it a standalone tool that people can access on the web by going to imagine.meta.com.
The company is calling this new standalone tool Imagine with Meta AI. To use it, all you have to do is visit the website mentioned above, describe the image that you want the tool to generate in the text format in the text box, and click on the Generate button. Before processing the first request, the website will ask you to log into the platform using a Meta account. Once you do that, it will generate four images based on the description you provided. At the moment, Imagine is only available in the US.
Meta used 1.1 billion images to train Imagine
Imagine with Meta AI is powered by Emu, which is Meta’s “image foundation model.” In a blog post, the company explains that it has trained Emu using 1.1 billion images. According to a report from Reuters, Meta used publically available images on Facebook and Instagram to train the image generation model. Imagine will add a watermark at the bottom-left portion of the images it generates to indicate that those pictures were generated using AI. Meta will soon update Imagine to add invisible watermarks.
