Pages

Meta’s New AI Image Generation Engine Hopes To Create The Best Digital Art And Immersive Worlds

It’s to surprise that AI-powered technology is ruling the digital world and that’s one of the many reasons why so many tech giants have adopted them into their platforms.

These days, Text-Image production is trending as the leading algorithmic endeavor. We’ve got some notable examples in front of us like Google’s Imagen which is proving that AI is here to stay, thanks to its incredible results.

The sublime blend of human and computer creativity has allowed for the creation of the best digital art. And that’s one of the reasons why Meta has worked hard in playing its role toward designing the best AI-based image production engine.

The hopes are plenty and the end target goals are also quite high from this new machine through which Meta hopes to get fabulous results in terms of digital art and also the immersive world.

The leading tech company announced on Tuesday that it had created a new AI engine to assist in its metaverse and more.


For those who don’t know, there’s a lot of work that goes into producing the right image from a simple text or phrase.

Firstly, you make the phrase enter the transformer model. The latter is a network that puts together several words in one sentence and then tries to adopt some sense into what it really means in regards to the words together.

But once the technology gets the hang of what is actually being described, the engine words to create a new picture with the help of GANs which is the term reserved for generative adversarial themed platforms.

Remember, AI didn’t just magically enter into our lives. We’ve really seen it come a long way from where it once began. For years, ML models used to undergo vigorous training sessions upon image sets of the finest quality.

This coupled with the best text descriptions gives rise to modern-day AI gadgets whose end results are nothing less than the best and most realistic images that one can imagine. At the end of the day, it’s mostly dependent on what you give them and hence what they end up producing.

The only difference between one AI engine with another has to do with the type of creation process involved.

Let’s take Google’s Imagen into consideration. The device makes use of dots placed randomly and turns them into a glorious picture. With time, the resolution increases from low at first to higher with time.

Others could potentially think about converting a particular image into several code entries and then assembling those to give the end product.

So as you can see, the change varies from one product to another. But one thing in common is that any user has no say in how the final image will turn out.

This is exactly why Meta’s CEO revealed through a recent blog post how the old approach behind these AI engines was to see their true potential and what they’re capable of. But now, he hopes that new advancements would allow humans to have a say and control what type of image is being generated.

Meta has called its new AI design engine an exploratory research endeavor and it’s even gotten the nickname ‘Making a Scene’.

Meta revealed how it plans on feeding user-generated sketches to the models so that a particular dimension image is produced in a certain pixel. Well, that definitely sounds very unique and customized to us.

This way, users will have so much control of what’s being produced and they can also now dictate what its composition will comprise.

It’s quite clear that Meta’s goals are more related to conveying a particular vision with so much more intricate detail and specificity.

Read next: Facebook Users Will Soon Be Allowed To Create Up To Four Additional Profiles

No comments: