Published on December 7, 2023, 9:14 pm
Meta AI Launches Imagine: A Free Image Generator Powered by Generative AI
In the midst of Google’s Gemini launch, Meta has unveiled Imagine, a standalone image generator website fueled by Meta AI. This technology has been in development for some time, with Meta experimenting with it through a beta version on WhatsApp since August of this year (as reported by TechRadar). Previously, accessing the feature required the Meta app to be installed on smartphones. However, with Imagine, all that is needed is an email address to create an account and start utilizing this powerful generative AI platform. Think of it as similar to OpenAI’s DALL-E.
Curiosity piqued, we decided to test out the website ourselves. What we found was that the AI is capable of generating four JPEG images sized at 1,280 x 1,280 pixels. To download them, simply click on the three dots in the upper right corner and select your preferred option from the drop-down menu.
We were pleasantly surprised to discover that Imagine can create content featuring well-known cartoon characters like Homer Simpson and even Mickey Mouse. One would expect copyright restrictions to prevent such usage, but apparently not. However, while these generated images are impressive, they do come with noticeable flaws. For example, in our sample of Homer Simpson, parts of the picture appeared distorted or melded together. Additionally, the character itself looked somewhat peculiar.
Limitations and Workarounds:
The development of Imagine involved significant care and was made possible by Meta’s proprietary Emu learning model. In a research paper published in September (AI.Meta), it was revealed that Emu had been trained using an astonishing “1.1 billion images.” Although the source of this vast dataset had previously been undisclosed to the public, Nick Clegg (Meta’s president of global affairs) confirmed to Reuters that public Facebook and Instagram posts were used for training purposes. In total, data was scrapped from over a billion social media accounts.
To manage this massive dataset, Meta has implemented some restrictions. The AI will refuse prompts that are violent or sexual in nature, as well as those mentioning famous individuals. Despite their best efforts, however, the technology is not flawless. It seems that users can bypass these limitations by using indirect wording. For instance, when we attempted to create an image of former President Barack Obama directly, the AI refused. However, when we rephrased the prompt to “a former US president,” the AI generated an image resembling President Obama.
Meta has plans to introduce “invisible watermarking… for increased transparency and traceability,” but this feature is still a few weeks away from being released. In the meantime, concerns about potential misuse remain. We reached out to Meta regarding their intentions to implement additional protection measures and will update this story accordingly.
If you’re interested in exploring the world of AI art generators further, be sure to check out TechRadar’s comprehensive guide on the best options available in 2023.
You might also like:
– Facebook Messenger gets its biggest ever update – including a major privacy boost
– WhatsApp beta introduces AI assistant to chats – here’s how it works
– How much more data can Meta collect? Probably a lot, thanks to the Meta Quest 3 and Ray-Ban smart glasses
In conclusion, Meta’s Imagine opens up exciting possibilities for creative expression through generative AI-powered image generation. With its intuitive interface and accessibility, anyone can now explore their artistic side with just an email address. However, as with any new technology, there are limitations and potential concerns that need addressing moving forward. Nonetheless, Imagine represents another step forward in the ever-evolving field of artificial intelligence and its applications in our daily lives.
Images Courtesy of Future